~~Speaker: Assaf Peled, 
M.Sc. student under the supervision of Prof. Joseph Appelbaum

Wednesday, March 4, 2015 at 15:30
Room 011, Kitot Bldg., Faculty of Engineering

Modeling Aspects in Photovoltaic Devices and Their  Environmental Factors

Abstract

The thesis study is organized in three sections which encompass topics from the fields of solar energy and photovoltaics. Thematically, the study first discusses how environmental factors affect PV systems, and then works its way down into the solar-cell level.
The first part concerns the energy production's intermittency of outdoor PV facilities due to stochastic movement of clouds. In this part, an analysis was conducted on global solar irradiance time-series under clear and overcast weather conditions. One effective method for isolating and identifying the underlying patterns of solar fluctuations is the use of wavelets and wavelet transform. A cross-correlation between a wavelet's waveform and a solar irradiance time signal produces a value which quantifies how similar those two functions are. Through a careful analysis, it is shown that although an outdoor PV system is frequently subjected to environmental changes, its power yield can be effectively predicted.
A novel solar spectral irradiance model based on artificial neural networks (ANN), is established in the second part of the thesis. The ANN typically starts with a learning phase, in which it learns to recognize persisting input-output relations. The model's output is gradually refined so as to minimize the error between the real-measured output data and the modeled outcome. Separate neural-networks are designated to simulate the spectral solar irradiance at a specific wavelength. The complete solar irradiance spectrum is assembled from all individual fractional spectra. The spectral model is further tested against an empirical measured spectral data, showing a remarkable performance under all-sky conditions.
In the third part of the thesis, methods to extract the electrical parameters of solar cells are presented based on the single-diode model and on the cell's I-V curve. Three different optimization methods are employed. The study aims to emphasize the discrepancies between the extraction's methodologies in terms of accuracy and time efficiency. All methods are tested against real measured parameter values of GaAs and a Multi-Junction cells.

• מצגת 1
• מצגת 2
• מצגת 3
• מצגת 4
• מצגת 5

~~Hadas Aharon, 
M.Sc. student under the supervision of Prof. Anthony J. Weiss

Wednesday, February 25th, 2015 at 15:00
Room 011, Kitot Bldg., Faculty of Engineering

Urban Localization Using Ray-Tracing
Target localization is one of the most interesting and important applications offered by wireless networks. Location based services are increasingly attracting the attention of service providers as well as of researchers. The traditional location methods, such as time of arrival (TOA), angle of arrival (AOA) and their derivatives, are widely used in the fields of emitter localization and vehicle navigation. These traditional methods, and even the newer method - direct position determination (DPD), are based on the line of sight (LOS) assumption. As time goes by, the landscape of more populated areas is filling up with buildings, especially tall ones, and the change in the landscape makes the LOS assumption invalid. The loss of LOS causes a significant decline in the accuracy of the above mentioned methods.
A natural direction for finding the position of an emitter in an unknown wave propagation environment is to espouse the fingerprinting approach, which is what we did. In the fingerprinting approach, features of signals received from a collection of discrete transmission points, are kept in a database. Every time a signal is received from a mobile terminal in an unknown location, it is compared with the information in the database regarding each of the discrete transmission points, and the estimated mobile terminal's location is selected among these points. In this thesis we used predictions made by a ray-tracing tool, provided by NICE Systems LTD., to build our database.
This research presents several solutions to the problem of localization in urban environments. First, we solve the problem under a naive assumption about the channel attenuation coefficients; then, we refine our assumption and solve the problem again. For both methods we offer two solution schemes - a direct scheme and a two-step scheme, where the direct scheme detects the transmission location using the Least-Squares (LS) criterion and the two-step scheme adds a preliminary stage designed to reduce the computational complexity. For the purpose of developing the preliminary stage, we analyze the output of the LS cost function, and from this analysis, we quantify the effect of different model orders of the various options, on the selection of the detected transmission point.
Tests of our solutions, using simulations of signals transmitted by a GSM cellular phone from unknown locations within an area of 88mX650m, indicate that locating a cellular phone with accuracy of less than 20m in 76.1% of the cases, is possible.