UPM Institutional Repository

Unified neural network controller of series active power filter for power quality problems mitigation


Citation

Ghazanfarpour, Behzad (2013) Unified neural network controller of series active power filter for power quality problems mitigation. Masters thesis, Universiti Putra Malaysia.

Abstract

This thesis presents the design, development and analysis of series active power filter (APF) with a novel control technique. The series APF developed in this work is applied at a point of common coupling (PCC) to compensate voltage harmonic and source-end disturbances of the grid supply voltage. To solve the mentioned power quality issues, the architecture of the series APF controller is developed to assess and extract voltage harmonic, sag, swell and interruption conditions. The proposed unified controller is responsible to generate proper switching signals for dynamic compensation of voltage harmonic, sag, swell and interruption and formed of two main parts. The first and core unit of the controller utilizes adaptive neural network algorithm to extract fundamental component of the supply voltage and detach the present voltage harmonic. The adaptive neural network unit uses Adaline structure for faster extraction of present distortions at the supply voltage. The amplitude value of the fundamental component is measured and saved by the second unit called peak detector. Furthermore, it is unified with a phase-locked loop (PLL) based reference generator unit to generate proper compensation signal for present faults at voltage supply fundamental signal caused by source-end disturbances. In order to enhance the response time of series APF, two of the most well known learning algorithms are investigated in this work. First, Widrow-Hoff algorithm is examined and its constant learning rate is modified by adding an adaptive learning rule to change the learning rate value. This modification demonstrates a significant contribution in performance of the series APF in terms of speed and accuracy. However, in the second investigation neural network unit utilized Levenberg-Marquardt algorithm as the harmonic extractor, since it is faster and more accurate than adaptive Widrow-Hoff especially during fundamental signal variation caused by source-end disturbances. The controller uses extracted distortion signal to generate the proper reference signal. The compensator unit receives reference signal to operate the pulse width modulation (PWM) based voltage source inverter to generate the replica of the voltage harmonic, sag, swell and interruption to inject them into the system for compensation. The implementation of the proposed APF could maintain the utility supply voltage at the point of common coupling in a distribution network at almost sinusoidal and should minimize the total harmonic distortion (THD) levels in the network. The proposed extraction techniques and inverter control scheme of series APF topology was investigated to examine the functionality of the system under different disturbance conditions. The obtained simulation results illustrate the capability of the series APF with the proposed unified controller in mitigating of voltage harmonic and compensating of voltage sag, swell and interruption during variety of voltage quality problems occurrence at the PCC.


Download File

[img]
Preview
PDF
FK 2013 68R.pdf

Download (1MB) | Preview

Additional Metadata

Item Type: Thesis (Masters)
Subject: Electric power systems - Control
Subject: Neural networks (Computer science)
Call Number: FK 2013 68
Chairman Supervisor: Mohd Amran Mohd Radzi, PhD
Divisions: Faculty of Engineering
Depositing User: Haridan Mohd Jais
Date Deposited: 22 Jul 2016 04:36
Last Modified: 22 Jul 2016 04:36
URI: http://psasir.upm.edu.my/id/eprint/47591
Statistic Details: View Download Statistic

Actions (login required)

View Item View Item