Seminars and Defenses

All are welcome and encouraged to attend the seminars and defenses.

Jan. 8, 10AM, ENG460
Saman Alaeddini • MASC thesis defense
Holistic approach to achieve wide area protection coordination
The electric power transmission network of today is undergoing significant changes in terms of the operational requirements, connected distributed energy resources technologies, and regulatory requirements. In conjunction with an aging infrastructure, these changes have presented new challenges to utilities in their fundamental mission of providing reliable electrical power to the customers. Thus, protection systems must overcome additional challenges towards safe and reliable operation of the power system. Increased investigation of protection system performance is therefore needed to ensure proper coordination of protective relays. However, the complex and integrated nature of the modern protection and control systems call for more sophisticated modeling and study tools for the simulation and analysis of both the dynamics of the interconnected transmission systems and interactions among numerous sets of intelligent electronic devices. Although the protection technology designed for transmission systems is mature, coordination of devices is still a major challenge. This thesis presents a holistic approach to conducting wide-area protection coordination studies through the use of a practical automation-assisted methodology. The wide area protection coordination solution covers process and data management considerations. It also provides a framework workflow for the execution and review of coordination studies, as well as processing and documentation of results to support reliability improvements. The fundamental concept behind the proposed approach is the utilization of software-based automation in a number of key tasks. Firstly, the execution of large-scale protection system coordination studies can be largely automated through utilization of specialized scripts running within short circuit simulation software packages. Secondly, the vast amounts of data inherent in the protection system coordination study results are processed in a manner that assists protection engineers in the identification and resolution of coordination issues. Finally, user friendly automated study summaries are generated that can be used as a record of protection setting recommendations. The effectiveness of the proposed solution is demonstrated through simulations conducted in the CAPE software environment for short circuit studies.
Jan. 7, 10AM, ENG460
Jahan Afsharian • PHD thesis final defense
High power density matrix converter based power supplies for telecom and datacenter applications
With the fast development of information technology (IT) industry, the demand and market volume for off-line power supplies keeps increasing, especially those for telecommunication, computer server and datacenter. The total power consumption of today's data centers is = becoming noticeable. In 2014, data centers in the U.S. consumed an estimated 70 billion kWh, representing about 1.8% of total U.S. electricity consumption. Moreover, with the increase in cloud computing and big data, energy use of data centers is expected to continue rapidly increasing in the near future. Generally, the power supply (PSU) for power distribution system (PDS) in datacenter and telecom are the standard two-stage approach. An off-line PSU normally consists of power factor correction (PFC) circuit and isolated dc-dc converter. The front-end PSU needs to meet several rigid requirements of IT equipment, such as input to output isolation, tight output voltage regulation, high-efficiency, high-power-density and n+1 parallel operation. Besides these indispensable requirements, power quality is a major concern and stringent international requirements, such as the IEC 61000-3-2, is enforced to limit the harmonic currents drawn by the off-line equipment. The PSU has to meet the requirements for abnormal input operation such as loss of one phase in three-phase operation, voltage sag, unbalanced input voltage and surge voltage. In general, the two-stage power conversion has demonstrated excellent performance and high reliability, since the design can be optimized for each stage. In order to fulfill future requirements for the PDS in datacenter and telecom applications, it is therefore of great importance to identify new ways to develop systems with higher efficiency, power density, reliability and lower cost. A very promising and fundamentally different approach, in order to fulfill the future demands for PDS in datacenter and telecom, is the development of single-stage converter. The development of single-stage matrix-type converter creates new degrees of freedom regarding e.g. simplified rectifier racks in telecom and datacenter. This provides tangible benefits in the form of space saving, and better airflow for power unit in rectifier racks.
Dec. 3, 10AM
Amr Adel Fathy Mohamed • PHD internal exam
Line-wise power balance equations and applications
Optimal power flow (OPF) is the optimization of a chosen objective function by determining the best values of control variables while satisfying a set of equality and inequality constraints. Even though commercial OPF algorithms are robust and efficient, they still cannot guarantee a global optimum. The US Federal Energy Regulatory Commission estimates that the best commercial OPF solvers are off by 10%, precipitating an economic loss of US $400 billion per year worldwide. In addition, the continuous necessity to ensure voltage stability of power systems while operating at the minimum cost necessitates incorporation of voltage stability constraints into optimal power flow formulations. For these reasons and given the nonlinear and nonconvex characteristics of the optimization problem, there is significant technical and economic impetus whereby OPF remains a major research focus. This thesis aims to (1) develop a new set of power balance equations yielding faster power flow (PF) algorithms with better solution space that are the best suited for OPF applications, (2) build new OPF formulations using this new set of power balance equations, and (3) incorporate voltage stability constraints into the newly developed OPF formulation. The genesis of the new set of power balance equations stems from the fact that power of a constant impedance load is proportional to the square of voltage magnitude. Using square of voltage magnitude in conjunction with angles of bus voltage phasors, a set of line-wise power balance equations is developed, both in polar and rectangular forms. The solution algorithm of these two formulations is developed using Newton-Raphson (NR) technique. All developed methods are tested on 6-, 14-, 57- and 118-bus IEEE systems, a 582-bus real system, a 2383-bus Polish power system, and a 9241-bus PEGASE system. Tests show that the developed line-wise power flow (LWPF) is accurate, provides monotonic convergence, and scales well for large systems, while using sparse matrices. Studies also show that the newly proposed LWPF methods are up to twice and thrice faster for polar and rectangular forms respectively, when compared with conventional bus-wise power flow (BWPF) methods that use bus-wise power balance equations. Numerical analysis of the Jacobians of BWPF and LWPF methods shows that the number of calculations required for the LWPF method is much lower, resulting in the higher speed. Another significant benefit of the polar form LWPF method is the ability to directly identify the set of critical lines that connect busses that are the most susceptible to Voltage collapse (VC). Based on the polar form of a line-wise set of power balance equations, a novel line-wise optimal power flow (LWOPF) is developed. LWOPF is linearized and solved using successive linear programming. Due to the better solution space modeled by LWPF equations, the optimal solution yielded by the LWOPF is consistently better than that computed using bus-wise OPF. The polar form of LWOPF is extended to include voltage stability constraints and implemented considering both linear and nonlinear optimization techniques. It demonstrates improved performance in achieving lower cost optimal solutions with better voltage-stable states.
Nov. 16, 11:00AM, EPH 111
Prof. Gene Cheung • Research Seminar
Graph Spectral Image Processing
The emerging field of graph signal processing (GSP) has focused on signals that live on irregular data kernels described by graphs.  Recently, GSP techniques have also been effectively applied to imaging data that reside on 2D grids. We focus the talk on recent advances in two application areas of GSP for imaging: i) compression, and ii) restoration. For compression, we overview the derivation and use of graph Fourier transform (GFT) and its variants for image compression. Application extends to modern image modalities, such as point cloud and light field images. For restoration, we discuss the use of graph Laplacian regularizer (GLR) and graph total variation (GTV) for image denoising and deblurring of natural images. Application extends to point cloud denoising and inpainting. Finally, we discuss future direction of unrolling graph-based iterative algorithms into neural net layers for efficient implementation and robustness guarantees for inverse imaging problems.
Gene Cheung received the B.S. degree in electrical engineering from Cornell University in 1995, and the M.S. and Ph.D. degrees in electrical engineering and computer science from the University of California, Berkeley, in 1998 and 2000, respectively. He was a senior researcher in Hewlett-Packard Laboratories Japan, Tokyo, from 2000 till 2009. He was an associate professor in National Institute of Informatics in Tokyo, Japan till July 2018. He is now an associate professor in the department of EECS, York University, Toronto, Canada. He has been an adjunct associate professor in the Hong Kong University of Science & Technology (HKUST) since 2015. His research interests include 3D imaging and graph signal processing. He has served as associate editor for multiple journals, including IEEE Transactions on Multimedia (2007--2011), IEEE Transactions on Circuits and Systems for Video Technology (2016--2017), and IEEE Transactions on Image Processing (2015--present). He served as a member of the Multimedia Signal Processing Technical Committee (MMSP-TC) in IEEE Signal Processing Society (SPS) (2012--2014), and a member of the Image, Video, and Multidimensional Signal Processing Technical Committee (IVMSP-TC) (2015--2017, 2018--2020). He is a co-author of many paper awards in SPS, including the best student paper award in IEEE ICIP 2013, ICIP 2017 and IVMSP 2016, best paper runner-up award in ICME 2012, and IEEE SPS Japan best paper award 2016.
Nov. 21, 11AM, ENG471
Noha Hassan • PHD final defense
Finding Optimal Macro and Micro Base Stations Locations and User Associations in Heterogeneous Wireless Networks
Massive Multiple Input Multiple Output (MIMO) systems have gained the attraction of the communication industry recently, due to its promising ability to out peak the performance of future broadband Fifth Generation (5G) networks. Also, Heterogeneous Networks (HetNets) are integral part of 5G systems. They can be viewed in multi-dimensional space where, each slice represents a unique tier that has its own Base Station (BS)s and User Equipment (UE)s. Different slices cooperate with each other for mutual benefit. Data can be interactively exchanged among the tiers, and UEs have the ability to switch between the tiers. The cells in such a heterogeneous cellular networks can no more be considered of as fixed size with a hexagonal shape.
Pilot sequences are used to acquire the Channel State Information (CSI) of the channel. Pilot sequences replacing data sequences has an adverse effect on channel spectral efficiency and throughput. Having long pilot sequences would reduce pilot contamination at the expense of spectral efficiency and throughput while too short pilot sequence would affect the orthogonality and channel estimation. In this thesis, we introduce analytical derivations for optimum pilot length to minimise mean squared error and maximise pilot power.
Poisson Point Process (PPP) has been widely used to allocate BSs and UEs among various tiers so far. However, BS locations using PPP approach may not be optimum to reduce interference. Based on this, we provide an algorithm to optimise Micro BS locations in high demand dense networks to minimise interference from adjacent BS's and improve UE coverage. Also, we formulate expressions for static UE's coverage probability and network energy efficiency in HetNets. Furthermore, we suggest Dummy Intersecting Circle (DIC) algorithm to optimise Macro BS locations and to completely eliminate all sources of interference from adjacent Macro and Micro BS's, then we introduced the modified version of the aforementioned algorithm Modified Dummy Intersecting Circle (MDIC) to provide a combined approach to optimise Macro BS and core Micro BS locations. Our algorithms proved to be feasible and provide a path towards attainable future communication systems.
Sep. 20, 12 Noon, ENG106
Dr. Fred Tung • Research Seminar
Deep Neural Network Compression
Deep neural networks enable state-of-the-art accuracy on visual recognition tasks such as image classification and object detection. However, modern deep networks contain millions of learned connections, and the current trend is towards deeper and more densely connected architectures. A more efficient utilization of computation resources would assist in a variety of deployment scenarios, from embedded platforms with resource constraints to computing clusters running ensembles of networks.
In this talk, I will first consider the common scenario of adapting a pre-trained neural network to a narrower, specialized image domain. I will introduce the fine-pruning method, which jointly fine-tunes and compresses the pre-trained network to produce an efficient network tailored to the target domain. Next, I will consider the general scenario of compressing a pre-trained neural network. I will present the CLIP-Q method (Compression Learning by In-Parallel Pruning-Quantization), which performs weight pruning and quantization jointly, and in parallel with fine-tuning. CLIP-Q compresses AlexNet by 51-fold, GoogLeNet by 10-fold, ResNet by 15-fold, and MobileNet by 7-fold, while preserving the uncompressed network accuracies on ImageNet. Finally, I will consider the scenario of obtaining a compressed neural network that satisfies known operational constraints. I will present constraint-aware network compression, which incorporates operational constraints directly in the compression process.
Fred Tung is an NSERC postdoctoral fellow working with Greg Mori at Simon Fraser University. His research interests are in computer vision and machine learning, with a focus on developing algorithms for training resource-efficient deep neural networks. He received the Ph.D. degree in computer science from the University of British Columbia under the supervision of Jim Little.
Sep. 13, 2PM, ENG471
Ishmeen Sra • MASC Thesis Oral Exam
CN Tower Lightning Flash Characteristics Based on High-Speed Imaging
In 1991, a significant phase of studying CN Tower lightning characteristics commenced, when five measurement systems operated simultaneously to record all relevant lightning parameters, including the optical characteristics, the lightning current and the corresponding lightning-generated electromagnetic pulse. This thesis emphasizes the analysis of fifty-eight flashes that struck the CN Tower during the last five years (2013-2017) based on the records of Phantom v5.0 digital high-speed camera, operating at 1 ms resolution, and a continuously operating Sony HDR PJ790VB digital camera. The luminosity variation with time for every CN Tower flash was carried out precisely for the characterization and statistical analysis of flash components. Although the characteristics of each storm configuration, which includes inter-flash times, are determined, the detailed analysis of each flash was analyzed, including flash duration, initial-stage current, flash multiplicity, inter-stroke time duration and the continuing current. Furthermore, an extensive comparison between two CN Tower major storms has been performed. The first took place on September 5, 2014 and the other that took place almost three years later (September 4, 2017). The analysis of CN Tower lightning flash characteristics is fundamental for understanding tall-structures lightning, which leads to the protection of tall structures against lightning hazards.
Sep. 10, 12 Noon, ENG106
Dr. Ping Wang • Research Seminar
Ambient Backscatter: A New Approach to Improve Network Performance for RF-Powered Cognitive Radio Networks
Cognitive radio is an intelligent radio network which aims to utilize spectrum more efficiently. With the development of radio frequency (RF) energy harvesting techniques, a new type of network has been introduced, i.e., an RF-powered cognitive radio network (CRN). However, the performance of RF-powered CRNs depends largely on the amount of harvested energy and the primary channel activity. Recently, ambient backscatter communication has been introduced, which enables users to transmit data by backscattering ambient signals. In this talk, we will introduce a novel concept of integrating ambient backscatter communication into RF-powered CRNs with the aim of improving the performance of the secondary system in terms of the overall data transmission rate. An important question to answer is how to choose the best mode, i.e., the harvest-then-transmit mode or backscatter mode, given the current radio conditions. To this end, we formulate and solve optimization problems to find an optimal transmission policy for the secondary system. Moreover, we analyze and evaluate the performance of the secondary system under different scenarios. Through the numerical results, we demonstrate that by integrating ambient backscatter communication into the RF-powered CRN, the performance of the secondary system can be significantly enhanced.
Sep. 12, 10AM
Nikoo Kouchakipour • MASC Thesis Oral Exam
A DC-Side Fault-Tolerant Bidirectional AC-DC Converter for Power System Integration of Low-Voltage DC Distribution Systems
With the rising potential for the employment of low- and medium-voltage direct-current (dc) electric power distribution systems, most notably for a more efficient integration of plug-in electric vehicles and such other distributed energy resources as photovoltaic (PV) panels, there is a need for robust ac-dc electronic power converters that can interface such dc distribution systems with the legacy alternating current (ac) power system. Thus, this thesis proposes a new single-stage low-voltage three-phase ac-dc power converter that is simple structurally, enables a bidirectional power exchanges between the ac and dc distribution systems, and can handle short-circuit faults at its dc as well as ac sides. The proposed converter consists of three legs, corresponding to the three phases of the host ac grid, each of which hosting two full-bridge submodule (FBSM), in an architecture that can be regarded as a special case of the so-called modular multi-level converter (MMC). Thus, the dc port each FBSM is connected in parallel with a corresponding capacitor, while the ac voltage of each phase is synthesized by the coordinated sinusoidal pulse-width modulation (SPWM) of the two corresponding FMSMs. This architecture allows the generation of low-distortion ac voltage while it also provides the converter with the very important dc fault current blocking capability since, upon the detection of a short circuit across the converter dc port, the switches of the FBSMs are turned off and disallow the flow of any dc current. The thesis also presents a mathematical model for the converter, for analysis and control design purposes. Thus, the control for the regulation of the overall dc-side voltage, as well as those for the regulation of the dc voltages of the FBSMs are devised based on the aforementioned mathematical model and presented with details. It is further shown that the voltage conversion ratio of the proposed converter is the same as that offered by a conventional voltage-sourced converter (VSC), whereas the VSC is vulnerable to dc-side shorts. The proposed converter can be extended to medium-voltage levels by multiplying the number of FBSMs in each leg. The effectiveness of the proposed converter and its controls is demonstrated through time-domain simulation studies conducted on a topological model of the converter in PSCAD/EMTDC software environment.
Sep. 12, 11AM
Heng Zhao • MEng Project Oral Exam
ANN-Based Day-Ahead Short Term Load Forecasting
Load forecasting (LF) is of great significance for effective operation, utilization, safety and reliability of the modern electric power systems. Load forecasting can be categorized into very short term, short-term, medium-term, and long-term forecasts, depending on which time scale is concerned. The short term load forecasting (STLF) plays an increasingly important role in achieving a more efficient, reliable and safe power system. Its outputs are the indispensable inputs of generating scheduling, power system security assessment and power dispatch. In the era of smart grid (SG), STLF is the basic building block to imply Demand Side Management (DSM) in areas such as automatic generation control, load estimation, energy purchasing, and contract evaluation, etc. The accuracy of STLF is of essential importance for both economic and reliability.
In the last few decades, various methods have been devised and applied to perform STLF. Due to its superior capability of handling the nonlinearity, Artificial Intelligence (AI) based techniques are gaining more popularity in a variety of applications. The objective of this study is to review, categorize, evaluate, and analyze the principle, application, and performance of STLF techniques. It builds up several feed forward Artificial Neural Networks (ANN) models with different configurations, and studies the mechanism of ANN for effective STLF. With 12 years of hourly load and meteorological data sets of a section of the City of Toronto, the configurations are built up with different hidden layers, activating function, training algorithms and both un-normalized and normalized data to predict the day ahead STLF with satisfactory result achieved.
Sep. 11, 10AM, ENG460
Lakmini Perera • MASC Thesis Oral Exam
Recovery of Valuable Incompletely-Recorded CN Tower Lightning Return-Stroke Current Derivative Signals
Lightning is a captivating natural phenomenon but indisputably terrifying. Therefore, lightning studies have played an essential role in establishing safety regulations to protect lives and infrastructures. Among the many simulating functions that were utilized in the past for modelling the lightning return-stroke current, Heidler and the Pulse functions overcame certain limitations, including the time-derivative discontinuities. Incompletely-recorded current derivative signals represent another challenge in lightning research. This thesis proposes a double-term Pulse function that could be investigated with the double-term Heidler function for modelling the lightning return-stroke current. The time-derivative of both the Pulse and Heidler functions have been used to simulate the current derivative signals recorded on June 10, 1996. Some of these return-stroke signals exceded the maximum set level. Consequently, the double-term simulating functions were used to recover a large incompletely-recorded returnstroke current derivative signal. The R 2 fitting factor was used to evaluate the quality of each fitting to determine which simulating function is better suited to model and recover valuable return-stroke current signals.
Sep. 10, 10AM, CUE Conference Room
Mohammadreza Vatani • MASC Thesis Oral Exam
Linear Power Flow Analysis Method for AC-DC Electric Power Networks
AC-DC power systems have been operating more than sixty years. Nonlinear bus-wise power balance equations provide accurate model of AC-DC power systems. However, optimization tools for planning and operation require linear version, even if approximate, for creating tractable algorithms, considering modern elements such as DERs (distributed energy resources). Hitherto, linear models of only AC power systems are available, which coincidentally are called DC power flow. To address this drawback, linear bus-wise power balance equations are developed for AC-DC power systems and presented. As a first contribution, while AC and DC lines are represented by susceptance and conductance elements, AC-DC power converters are represented by a proposed linear relationship. As a second contribution, a three-step linear AC-DC power flow method is proposed. The first step solves the whole network considering it as a linear AC network, yielding bus phase angles at all busses. The second step computes attributes of the proposed linear model of all AC-DC power converters. The third step solves the linear model of the AC-DC system including converters, yielding bus phase angles at AC busses and voltage magnitudes at DC busses. The benefit of the proposed linear power flow model of AC-DC power system, while an approximation of the nonlinear model, enables representation of bus-wise power balance of AC-DC systems in complex planning and operational optimization formulations and hence holds the promise of phenomenal progress. The proposed linear AC-DC power systems is tested on numerous IEEE test systems and demonstrated to be fast, reliable, and consistent.
Aug. 27, 9:30AM
Hossein Mahlooji • MASC internal thesis defense
Temperature Insensitive Fiber Optic Pressure Sensor With a Pi-phase Shifted FBG on Microstructured Fiber
Light wave sensing technology has become a competitive choice for strain measurement in many applications such as structural health monitoring or machine condition monitoring. Such success cannot be achieved without some advantages like being lightweight, electrically non-conductive, electromagnetic field and harsh environment immune, relatively high sensitive to strain change and the compatibility with wavelength division multiplexing method to measure or monitor several stations with just one line of the fiber optic cable. However, the use of Fiber optic sensors in pressure measurement in gas and fluid media encounters some limitations such as high sensitivity to ambient temperature changes, low sensitivity to pressure, and the measurement delay in gas medium. In this work, we demonstrate both analytically and by experiments that an FBG pressure sensor, drawn on a microstructure fiber with two side holes in its cladding, can be used to measure pressure and temperature simultaneously. With this method, we can remove the ambient temperature changes effect on pressure measurements. The sensor has a -phase shifted FBG which intrinsically has a much narrower linewidth than the conventional FBGs and can efficiently improve the sensitivity of sensor to pressure changes. The microstructure fiber has two different refractive indices along their two principal axes caused by its birefringence. The FBG peaks in measuring spectra relating to two principal axes change with different rates when the pressure and/or temperature applies which makes it possible to measure the change of pressure and temperature at the same time. Our results also show that our sensor responds to the pressure change instantaneously if the separation of two FBG polarization peaks is used as the measurand.
Sep. 10, 10AM
Mohammed Baljon • PHD final defense
Resource allocation for energy harvesting communication systems over fading channels under delay constraint
Energy Harvesting (EH) is an emerging communication paradigm to defeat the limitation of network longevity by recharging the nodes by harvesting energy from the environment. The Energy Harvesting Network (EHN) requires a stable and efficient power control scheme like other conventional communication systems. It is more complicated than that in conventional communication networks, in that it should not only consider the quality of service requirements of the network but also adapt the randomness of the energy arrival. In this thesis, several optimal offline and online resource allocation strategies for a point-to-point and two-hop EH communication networks over wireless fading channels are investigated.
As a first step, we introduce RGWF (Recursive Geometric Water-filling) algorithm that provides an optimal offline transmission policy for a point-to-point EH communication system. Next, we consider a network comprising of a source, a relay, and a destination, where the source is an EH node. Joint time scheduling and power allocation problems are formulated to maximize the network throughput by considering conventional and buffer–aided link adaptive relaying protocols. Based on the modified RGWF algorithm, we decouple the joint power allocation and transmission time scheduling problem and propose optimal offline and suboptimal online schemes. In the second part, we aim to obtain the optimal transmission policy that maximizes the average total throughput of a point-to-point EH communication system network in an online manner. The optimal solution is obtained using dynamic programming by casting the optimization problem as a semi-Markov decision process (SMDP). In a delay-tolerant approach, we consider a cross-layer adaptation, where the proposed policy chooses modulation constellation for EH networks dynamically depending on battery state, data buffer state in addition to channel state. The proposed SMDP-based dynamic programming approach has been approved in dynamically adaptive the change of the channel and/or buffers’ states that optimally satisfy the BER requirements at the physical layer, and the overflow requirements at the data-link layer.
Aug. 22, 3PM
Ghassan Mohammed Halawani • MEng project
Convolutional Neural Network for Image Classification Based on Transfer Learning Technique
The main purpose of this project is to design a convolutional neural network for image classification, based on deep learning framework. A transfer learning technique is used by MATLAB tool kit to modify and train the pre-trained network with a new dataset in order perform classifications of thousand images.
Firstly, the general common architecture of most neural network and its benefits are presented. The mathematical models and the role of each part in the neural network are explained in detail. Secondly, different neural networks are studied in term of architecture, application, and the working method to highlight the strengths and weaknesses of each of neural network. The last part conducts a detail study on one of the most powerful deep learning networks in image classification, convolutional neural network, and how it can be modified to suit different classification tasks by using transfer learning technique in MATLAB.
Aug. 22, 1PM
Adnan Arapovic • MEng oral exam
Fault Analysis Considering Fault Current Contributions from Distributed Energy Resources
For over a century, electric power systems used synchronous machines for the generation of electricity. However, with emerging concerns over climate change and the need for reduced greenhouse gas emissions, together with the growing awareness of the importance of the natural environment and the depletion of the earth’s non-renewable energy resources, the generation of electricity from distributed renewable energy resources such as solar photovoltaic (PV) and wind energy has begun to expand at a rapid pace. Proliferation of converter-based distributed energy resources at the transmission and distribution systems has introduced new challenges in determining the maximum possible fault currents that a power system must be able to withstand without being destroyed. As part of a connection impact assessment for a new generation facility, utilities require detailed information on fault characteristics of the generation sources. However, there is limited knowledge or contradictory conclusions regarding the behaviour of converter-based distributed energy resources (e.g., solar PV and wind power generation stations) during faults on transmission or distribution systems. Currently, concern is widespread throughout Ontario regarding the constraints limiting solar PV and wind power generation stations from connecting to the electrical grid for projects equal to or less than 500kW. Many solar PV and wind industry participants have had connection applications rejected or are experiencing delays regarding the connection of their projects. Therefore, it is imperative to develop the mathematical and software simulation models that approximate the response of converter-based distributed energy resources during a fault on the transmission or distribution system in order to determine the fault current contributions to the electrical grid that a transmission or distribution utility needs to reflect in their connection impact assessments. As part of this project, theoretical models that have been presented in various literary works and publications on the topic are implemented using a program developed in MATLAB for performing fault current analysis of a power system using conventional fault analysis techniques. The MATLAB fault current analysis program is used to perform fault analysis of a power system with converter-based distributed energy resources in order to demonstrate how to implement these models to approximate the fault response of converter-based distributed energy resources within the framework of conventional fault analysis techniques and accurately simulate the fault current contributions to the electrical grid that a transmission or distribution utility needs to reflect in their connection impact assessments.
Aug. 29, 11AM, ENG460
Abdullah Siddiqui • MASc thesis defense
Performance and Energy Optimization of Heterogeneous CPU-GPU Systems for Embedded Applications
One of the most critical steps of embedded system design is Hardware-Software partitioning. It is characterized by distributing the tasks or components of an embedded application between hardware and software such that the system-designer’s functional and non-functional constraints are satisfied. Heterogeneous computing platforms consisting of CPUs and GPUs have tremendous potential for enhancing the performance of embedded applications. The challenges of application partitioning for CPU-GPU mapping and scheduling are greater for such platforms due to their unique and diverse characteristics. In this thesis, an optimization methodology is devised and presented for partitioning and mapping computational tasks on CPU-GPU platforms while keeping a check on the power consumption. Our methodology also uses parallelism in applications and their tasks by utilizing the architectural capabilities of GPU. The optimizing algorithm is tested for a MJPEG decoder application as well as several benchmarks and synthetic task graphs.
Aug. 10, 11AM, ENG460
Hasan Farahneh • PhD final thesis defense
Investigation of Vehicle to Vehicle Communication System Using Visible Light Technology
Visible Light Communication (VLC) has emerged as an attractive alternative to radio frequency (RF) communication, due to cost-effectiveness and being license free. It has also proved its applicability in automotive applications, as an alternative and/or a complement to the traditional RF based communications. We investigate the suitability of VLC in Intelligent Transportation Systems (ITS) and discuss its advantages in terms of safety enhancement and improved efficiency of the ITS. In this thesis, a VLC-based Vehicle-to-Vehicle (V2V) system in practical environments, considering both Line-of-Sight (LOS) and Non-Line of Sight (NLOS) paths is presented. The thesis investigates robust communication between a Light Emitting Diodes (LEDs) based VLC emitter and Photodiodes (PDs) based VLC receiver. For consideration of a V2V communication system, we consider transmitter on vehicle headlights and receivers on taillight making (2x2) Multiple-Input-Multiple-Output (MIMO) communication link. A closed form expression of the Channel Impulse Response (CIR) is derived and the effect of various channel parameters is analyzed. Optical-Orthogonal Frequency Division Multiplexing (O-OFDM) with adaptive modulation schemes is proposed for system improvement. Its performance is evaluated in terms of Inter Symbol Interference (ISI) mitigation, and data rate improvement. Moreover, the effect of sunlight on V2V-VLC system is investigated with appropriate denoising schemes. Two denoising schemes are proposed and evaluated as a solution to combat the effect of the solar irradiance on VLC signal. Firstly, we use a differential receiver for denoising purposes followed by K-Nearest Neighbor (KNN) based adaptive filtering algorithm, which is a supervised Machine Learning (MLE) technique. The shadowing effect is also studied. Moreover, an application of VLC in Foglet based ITS is described. The simulation validation of the VLC-based V2V system is performed under various environmental conditions and scenarios. Obtained results emphasize the suitability of VLC technology for automobile applications.
July 30, 10AM, ENG460
Marcos Aguirre • MASC thesis defense
A Single-Phase DC-AC Dual-Active-Bridge Based Resonant Converter for Grid-Connected Photovoltaic Solar Applications
In the wake of the global energy crisis, the integration of renewable energy resources, energy storage devices, and electric vehicles into the electric grid has been of great interest towards replacing conventional, fossil-fuel-dependent energy resources. This thesis presents the circuit topology and a control strategy for a 250-W single-phase grid-connected dc-ac converter for photovoltaic (PV) solar applications. The converter is based on the dual active bridge (DAB) kernel employing a series-resonant link and a high-frequency isolation stage. For interfacing the 60-Hz ac grid with the 78-kHz resonant circuit, the converter utilizes a four-quadrant switch array that functions as an ac-ac stage. Therefore, a bipolar low-frequency voltage source, that is the grid voltage, is used to synthesize a symmetrical high-frequency voltage pulse-train for the resonant circuit. Thus, soft switching and the use of a compact ferrite-core transformer have been possible. Then, a fast current-control loop ensures that the converter injects a sinusoidal current in phase with the grid voltage, while a relatively slower feedback loop regulates the converter dc-side voltage, that is, the PV array voltage, at a desired value. To simulate the converter and to design the controllers, the thesis also presents nonlinear large-signal and linearized small-signal state-space averaged models. The performance of the converter is assessed through simulation studies conducted using the aforementioned averaged models, a detailed topological model in the PLECS software environment, and a prototype.
July 6, 3:30PM, ENG460
Sulaiman Aljeddani • MASC thesis defense
Runtime thermal management based on task migration techniques in 3D chip multiprocessors
The industry trend of Chip Multiprocessors (CMPs) architecture is to move from 2D CMPs to 3D CMPs architecture which obtains higher performance, more reliability, and reduced memory access latency. However, one key challenge in designing the 3D CMPs is the thermal issue as a result of maximizing the throughput. Therefore, applying Runtime Thermal Management (RTM) has become crucial for controlling thermal hotspots. In this thesis, two methods of run-time task migration are presented to balance the temperature and reduce the number of hotspots in the 3D CMPs. The proposed techniques consider hotspots both in the core and the memory layers simultaneously to make the optimum run-time task migration decisions. The first proposed approach is divided into two algorithms working in parallel, which aim at maximizing the throughput on the 3D CMPs while satisfying the peak temperature constraints. Experimental results show that the proposed architecture yields up to 60% reduction in overall chip energy. The proposed architecture improves the IPC for canneal and fluidanimate applications by 16% and 14%, respectively. In the second method, the proposed technique migrates the hottest core with the optimal coldest core instead of the coldest core in the core layer. The optimal coldest core is selected by considering hotspots DRAM banks in the memory layer. The simulation results indicate up to 33℃ (on average 13℃) reduction in the cores' temperature of the target 3D CMPs. Finally, the proposed techniques are efficiency clarified in the simulation results that the maximum temperature of cores in the core and memory layers are both less than the maximum temperature limit, 80℃.
Aug 20, 1PM, ENG460
Shima Mohammadali Pishnamaz • MASC thesis defense
Retinal Fundus Image Processing and Ensemble Learning: Optic Disc and Optic Cup Segmentation
Ophthalmologists have widely used retinal fundus imaging systems to examine the health of the optic nerve, vitreous, macula, retina and its blood vessels. Many critical diseases can be diagnosed by analyzing retinal fundus images. Retinal image-based glaucoma detection is a comprehensive diagnostic approach that examines the head cup-to-disc ratio (CDR) as an important indicator for detecting the presence and the extent of glaucoma in a patient. The accurate segmentations of the optic disc and optic cup are critical for the calculation of CDR. Therefore, machine learning based algorithms are very helpful to efficiently exploit the vast amounts of retinal fundus data.
In this thesis project, the main goal is to develop image processing and machine learning algorithms to automatically detect the optic disc and optic cup from fundus images. This goal has been achieved by developing and applying several image enhancement techniques. First, an algorithm is proposed and tested on several fundus images to detect the optic disc. The proposed algorithm is based on a combination of contrast limited adaptive histogram equalization, alternative sequential filters, thresholding, and circular Hough transform methods. It has been shown in the results section that the proposed algorithm is highly efficient in segmentation of optic disc from other parts of the fundus image. Several classification and modeling methods are studied in order to classify detected optic disc into optic cup and non-optic cup regions. In this thesis project three main ensemble modeling algorithms are studied to segment optic cup. The studied ensemble models are Random Forest (RF), Gradient Boosting Machines (GBM), and Extreme Gradient Boosting Machines (XGBoost). The comparison between these models shows that they have more accurate results than conventional classification methods such as Linear Regression (LR) or Support Vector Machines (SVM). This study shows that XGBoost is the fastest and most accurate approach to segment optic cup within the optic disc region.
May 4, 11AM, ENG471
Prathap Siddavaatam • Final PHD Thesis Defense
Today IoT is a major disruption that will mark an epoch in communication technology such that every physical object can be connected to the Internet. With the advent of 5G communications, IoT is in urgent need of optimized architectures that can efficiently support wide ranging heterogeneous multi-objective requirements of communication, hardware and security aspects. The optimization challenges are rooted in the technology and how the information is acquired and manipulated by this technology. Our research in this thesis provides a description of compelling challenges faced by IoT and how to mitigate these challenges by designing resource-aware communication protocols, resource-constrained device hardware with low computing power and low-powered computational security enhancements. This thesis lays the foundation for optimizing these challenging IoT paradigms by introducing a novel Delta-Diagram based synthesizing model. The Delta-Diagram provides a road-map linking the behavioral and structural domains of a given IoT paradigm to generate respective optimizer domain parameters, which can be utilized by any optimizer framework. The fundamental part of the communication synthesizer is a mathematical model, developed to obtain the best possible routing paths and communication parameters among things. The ultimate aim of the entire synthesis process is to devise a design automation tool for IoT, which exploits the interrelations between different layer functionalities. This thesis also proposes a novel cross-layer Grey wolf optimizer for IoT, which outperforms some of the contemporary optimizer algorithms such as Particle Swarm, Genetic Algorithm, Differential Evolution optimizers in solving unimodal, multi-modal and composition benchmark problems. The purpose of this optimizer is to accurately capture both the high heterogeneity of the IoT and the impact of the Internet as part of delta diagram synthesis enabled network architecture. In addition, the Grey wolf optimizer for IoT plays a crucial role in design exploration of system on chip architecture for IoT device hardware. The results generated by the optimizer yielded the most optimum feasible solutions in the design space exploration process of the IoT.
May 4, 11AM, ENG460
Yuguang "Michael" Fang • Seminar
Beef Up the Edge: How to Build a More Powerful IoT System
Connected things in various cyber-physical systems (CPSs), namely IoTs, enable us to sense physical environments, extract intelligent information, and better regulate physical systems we heavily depend on in our daily life. Unfortunately, how to design effective and efficient systems to meet specific applications with diverse quality of service requirements is of paramount importance but highly challenging due to the spatial and temporal variations of user traffic, network spectrum resource, computing capability, storage, and device types. One holistic design approach from the end-to-end perspective seems to be in dire need.
In this talk, the speaker will discuss various related problems and challenges in a connected world and then present a novel collaborative network architecture to enabling connected things to effectively harvest in-network capability (spectrum, energy, storage, and computing power) in a cognitive fashion and intelligently manage the spectrum efficiency, energy efficiency, and yes, security! By pushing in-network capability in communications, computing, and storage to the edge, this network architecture provides an effective and robust approach to IoT.

Dr. Yuguang "Michael" Fang received MS degree from Qufu Normal University, Shandong, China in 1987, PhD degree from Case Western Reserve University in 1994 and PhD degree from Boston University in 1997. He was an assistant professor in Department of Electrical and Computer Engineering at New Jersey Institute of Technology from 1998 to 2000. He then joined the Department of Electrical and Computer Engineering at University of Florida in 2000 and has been a full professor since 2005. He held a University of Florida Research Foundation (UFRF) Professorship (2006-2009, 2017-2020), a University of Florida Term Professorship (2017-2019) and Changjiang Scholar Chair Professorship awarded by the Ministry of Education of China (is currently affiliated with Dalian Maritime University).
Dr. Fang received the US National Science Foundation Career Award in 2001 and the Office of Naval Research Young Investigator Award in 2002, 2015 IEEE Communications Society CISTC Technical Recognition Award, 2014 IEEE Communications Society WTC Recognition Award, and multiple Best Paper Awards from IEEE Globecom (2015, 2011 and 2002) and IEEE ICNP (2006). He has also received 2010-2011 UF Doctoral Dissertation Advisor/Mentoring Award, 2011 Florida Blue Key/UF Homecoming Distinguished Faculty Award, and the 2009 UF College of Engineering Faculty Mentoring Award. He was the Editor-in-Chief of IEEE Transactions on Vehicular Technology (2013-2017), the Editor-in-Chief of IEEE Wireless Communications (2009-2012), and serves/served on several editorial boards of journals including IEEE Transactions on Mobile Computing (2003-2008, 2011-2016), IEEE Transactions on Communications (2000-2011), and IEEE Transactions on Wireless Communications (2002-2009). He has been actively participating in conference organizations such as serving as the Technical Program Co-Chair for IEEE INFOCOM’2014 and the Technical Program Vice-Chair for IEEE INFOCOM'2005. He is a fellow of the IEEE (2008) and a fellow of the American Association for the Advancement of Science (AAAS) (2015).
May 2, 10AM, ENG460
Md Forhad Ebn Anwar • MENG FINAL DEFENCE
Collision of vehicles in highways are very frequent. Because of high speed (more than 100 km/hour), the momentum of collision is too high that leads severe casualty. Automatic Driving Assistance system can assist the vehicle operators to take decision based on realistic practical calculation on safety measures. It is always better to have third eye working parallel with human to avoid road accident. There are several technologies used to develop perfect driving assistance system to achieve higher accuracy in detection, identification and distance measurement of obstacles where vision based system is one of them. Mono-vision system provides cheap and fast solution rather stereo vision. This project work conducted with objective to comprehend computational complexity in implementation of mono-vision camera based object detection where system will generate warning if the detected object has a motion towards target. Processing and analysing of captured video image is the focused mechanism of implementation and used internal image generator module to mimic actual video camera. Appeared size of the shape of object considered for the decision making.