Department of Electrical and Electronic Engineering
University of Peradeniya
Department of Electrical and Electronic Engineering
University of Peradeniya
The Sri Lankan Government promotes solar power through ‘Surya Bala Sangrahamaya’ and is hoping to develop 1000 MW of PV systems by 2024. Currently solar PV power systems in excess of 100 MW are in operation with roof-top solar modules making a small but significant contribution. However, in some areas, CEB is not allowed to connect roof-top PV modules due to the fears of possible over voltages in lines, overloading of lines and losing the supply security. Even though rooftop PV modules can be looked at as a burden to the last mile networks, if they are properly controlled and supervised they will bring many technical and economic advantages for the last mile networks.Therefore, in this research network management strategies executed through a Smart Distribution Management System (S-DMS) that integrates different controllable entities within the last mile network is considered to support these networks thus increasing the absorbability of rooftop PV. The management of the following controllable entities through a S-DSM in a coordinated manner will be investigated:Smart inverters of rooftop PV systems that can provide grid support through active and reactive power control,Smart meters connected to the consumer premises that can make the load flexible by controlling, shifting or switching off some smart and non-critical loads and charging and discharging plug in electric vehicles,Smart transformer at the origin of the last mile network that can change its secondary side voltage continuously to manage the network voltages.
: One of the main methods available to monitor fetal wellbeing is to monitor fetal movement patterns. It is one of the basic indicators of fetal health. Currently this is mainly done by the mother herself, which is very unstructured and time consuming. Therefore, the main target of this project is to come up with an end-to-end solution to monitor fetal movements reliably. An accelerometric sensor was developed and data were obtained. This data was highly contaminated with noise such as maternal movements, maternal laugh, maternal cough, etc. Therefore, several novel advanced signal processing algorithms were utilized to extract fetal movement signals. Initially algorithms such as Spectral clustering, Convoluted Neural Networks (CNN) and Non-Negative Matrix Factorization (NMF) based techniques were implemented. Currently more advanced machine learning algorithms such as Variable Auto encoders are being utilized.
Fractal antennas operate in multiple frequencies that depend on the number of iterations and the scale factor. However, the higher resonant frequencies are multiples of the fundamental frequency and therefore, the fractal antenna may not resonate at the frequencies the designer needs. Also the fractal antennas have no control over antenna characteristics such as the bandwidth and gain. This research project focuses on improving the performance of fractal microstrip antennas by modifying the fractals using genetic algorithms. Promising results have been obtained on improving the gain and bandwidth of Sierpinski carpet antennas with a less intensive genetic algorithm optimization process when compared with such a process required for an ordinary microstrip antenna.
A multispectral imaging system was developed which consists of nine spectral bands ranging from 400 nm - 1000 nm. The device was used to study the potential of multispectral imaging for food quality assessment. The device was also used to investigate the effect of contaminants in transformer insulators.
Following two studies were conducted with the collaboration of Faculty of Agriculture, University of Peradeniya.
Following study is an ongoing study with the collaboration of The National Aquatic Resources Research and Development Agency.
Following study is an ongoing study on utilizing multispectral techniques in insulator quality assessment.
One of the most important problems that insulators are subjected to is degradation, deterioration and contamination. The contaminants affect the external insulation performance of insulators and they reduce the effective creepage distance of an insulator. In order to avoid sudden electrical failures, insulators need to be properly maintained. Here the inventors proposed a method which utilizes the multispectral imaging techniques.
Remote sensing is the process of monitoring and detecting the physical characteristics of an area using reflected and emitted solar radiation at a distance. Typically satellites or aircrafts are used to acquire images using multispectral and hyperspectral cameras specially developed for remote sensing purposes. Owing to the ability to capture more spectral information, hyperspectral imagery is widely popular amongst the remote sensing community and in recent years an amalgam of applications and research has been proposed. The project was focused on two distinct aspects related to hyperspectral remote sensing: unmixing of hyperspectral images and mineral indication using hyperspectral images.
Hyperspectral unmixing (HU) is a crucial step in the hyperspectral image (HSI) analysis. It aims at decomposing the observed spectrum at each pixel into a collection of constituent endmembers, weighted by their abundances. For this several algorithms were proposed based on various premises such as signal and image processing, graph signal processing, and autoencoder architectures that are described below.
A novel blind HU algorithm referred to as Kurtosis-based Smooth Nonnegative Matrix Factorization (KbSNMF) which incorporates a novel constraint based on the statistical independence of the probability density functions of endmember spectra. Imposing this constraint on the conventional NMF framework promotes the extraction of independent endmembers while further enhancing the parts-based representation of data.
In a close spatial neighborhood, due to spatial autocorrelation, the abundances of an endmember tend to be similar to each other. Therefore, defining each abundance map as a signal on a suitable graph enables the HU problem to be analyzed for the first time from a graph signal processing perspective. This project introduces a novel Laplacian regularizer based on the l1-norm, where graph spectral analysis is utilized to show that the regularizer has natural piecewise smooth (PWS) signal promotion and noise rejection capabilities. A graph-based blind HU algorithm is developed by incorporating this regularizer and an l1/2-sparsity constraint into the NMF problem. Since the featured regularizer exploits the PWS property of abundance maps, the proposed method is effective in HU.
A novel architecture is proposed to perform blind unmixing in hyperspectral images. The proposed architecture consists of convolutional layers followed by an autoencoder. The encoder transforms the feature space produced through convolutional layers to a latent space representation. Then, from these latent characteristics, the decoder reconstructs the roll-out image of the monochrome image which is at the input of the architecture; and each single-band image is fed sequentially.
Mineral indication is the process of identification and classification of minerals using spectral images. The scope of the project was to develop an algorithm to identify and quantify the availability of limestone and establish the ground truths with the collaboration of the Department of Geology, University of Peradeniya, in the Northern region of Sri Lanka around the Jaffna peninsula
Where we are
Copyright 2015 - Department of Electrical & Electronic Engineering