Dissertations and Theses - Electrical and Computer Engineering
http://hdl.handle.net/2142/8888
Dissertations and Theses in Electrical and Computer EngineeringFast optimal power flow analysis for large-scale smart grid
http://hdl.handle.net/2142/46937
Fast optimal power flow analysis for large-scale smart grid
Liang, Yi
Optimal power flow OPF plays an important role in power system operation. The emerging smart grid aims to create an automated energy delivery system that enables two-way flows of electricity and information. As a result, it will be desirable if OPF can be solved in real time in order to allow the implementation of time-sensitive applications, such as real-time pricing. We develop a novel algorithm to accelerate the computation of alternating current optimal power flow (ACOPF) through power system network reduction (NR). We formulate the OPF problem based on an equivalent reduced system and then compute its solution. The detailed optimal dispatch for the original power system is obtained afterwards using a distributed algorithm. Our results are compared with two widely used methods: full ACOPF and the linearized OPF with DC power flow and lossless network assumption, the so-called DCOPF. Experimental results show that for a large power system, our method achieves 7.01× speedup over ACOPF with only 1.72% error, and is 75.7% more accurate than the DCOPF solution. Our method is even 10% faster than DCOPF. Our experimental results demonstrate the unique strength of the proposed technique for fast, scalable, and accurate OPF computation. We also show that the proposed method is effective for smaller benchmarks.
Smart Grid
Power System
Network Reduction
Optimal Power Flow
Thu, 16 Jan 2014 18:26:58 GMTA system for multi-level authentication with physically unclonable functions
http://hdl.handle.net/2142/46931
A system for multi-level authentication with physically unclonable functions
Konigsmark, Sven
We analyze deficiencies in existing Physically Unclonable Function (PUF)
systems and protocols, and propose a new system of PUFs (SoP) that is numerically
secure under extended attacker privileges and attack scenarios. Our proposed
system uses a multi-level authentication scheme and employs different designs of
PUF to achieve high security with low computational complexity and small
footprint.
By employing role-specific PUF designs, SoP reduces the area over existing
PUF-based authentication solutions by more than 68%. The key principles are: (i)
reduce assumptions required to guarantee numerical security to a minimum set of
practical assumptions; (ii) combine different PUF types to optimize security while
minimizing resource requirements; (iii) provide multiple layers of authentication as
a force-multiplier for the trusted party.
This multi-level protocol resolves security deficiencies with regard to manin-the-middle attacks and challenge-response-pair (CRP) storage issues in
conventional PUF protocols. Furthermore, SoP allows recognition and sealing of
security breaches. A mathematical formulation of the attack complexity and
statistical evaluation based on simulated PUF data show the strength of this new
protocol.
Physically Unclonable Functions
Hardware Security
Thu, 16 Jan 2014 18:26:47 GMTValue function approximation architectures for neuro-dynamic programming
http://hdl.handle.net/2142/46909
Value function approximation architectures for neuro-dynamic programming
Chen, Wei
Neuro-dynamic programming is a class of powerful techniques for approximating the solution to dynamic programming equations. In their most computationally attractive formulations, these techniques provide the approximate solution only within a prescribed finite-dimensional function class. Thus, the question that always arises is how should the function class be chosen?
In this dissertation, we first propose an approach using the solutions to associated fluid and diffusion approximations. In order to evaluate this approach, we establish bounds on the approximation errors.
Next, we propose a novel parameterized Q-learning algorithm. Q-learning is a model-free method to compute the Q-function
associated with an optimal policy, based on
observations of states and actions. If the size of a state or a policy space is too large, Q-learning is often not very practical because there are too many Q-function values to update. One way to address this problem is to approximate the Q-function within a function class. However, such methods often require an explicit model of the system, such as the split sampling method introduced by Borkar. The proposed algorithm is a reinforcement learning (RL) method, in which case the system dynamics are not known. This method is designed based on using approximations of the transition kernel of the Markov decision process (MDP).
Lastly, we apply the proposed results of value function approximation techniques to several applications. In the power management model, we focus on the processor speed control problem to balance the performance and energy usage. Then we extend the results to the load balancing and the power management problem of geographically distributed data centers with grid regulation. In the cross-layer wireless control problem, the network utility maximization (NUM) and adaptive modulation (AM) are combined to balance the network performance and transmission power. In these applications, we show how to model the real problems by using the MDP model with reasonable assumptions and necessary approximations. Approximations of the value function are obtained for specific models, and evaluated by getting bounds for the errors. These approximate solutions are then used to construct basis functions for learning algorithms in the simulations.
Neuro-Dynamic Programming
Parametric Q-learning
Value Function Approximation
Processor Power Management
Data Center Power Management
Cross-Layer Wireless Control
Thu, 16 Jan 2014 18:26:02 GMTMethods for increasing energy harvest with PV module integrated power converters
http://hdl.handle.net/2142/46898
Methods for increasing energy harvest with PV module integrated power converters
Ehlmann, Jonathan
Increasing energy harvest from PV systems and reducing overall PV system costs are important to the continued adoption of solar energy. In an effort to achieve this, power converters are being integrated into PV modules. These power converters allow for more flexibility in PV system design by enabling each PV module to operate independently. The physical integration of these power converters creates new possibilities to improve the energy harvest. This thesis proposes methods and circuits that leverage this physical integration to further enhance PV module integrated power converters.
A steady-state, SPICE-like, modified nodal analysis based (MNA) simulation tool is developed to study the effects of PV module mismatch and shading on PV systems. This simulation tool is used to calculate the power output of a PV system under varying conditions. This facilitates the comparisons of different algorithms and circuits for increasing energy harvest.
Certain assumptions can be made when it is known that a power converter will only be operating on a single PV module, and extra voltage nodes are accessible when a power converter is physically integrated into a PV module. Two global maximum power point tracking (MPPT) algorithms are proposed that leverage this fact to improve energy harvest.
Different circuit architectures are proposed to enable MPPT for a smaller subsection of PV cells called PV submodules. By tracking the MPP for PV submodules, further energy harvest improvements can be made. The proposed circuits include smaller power converters and multiple input converters for full PV submodule power processing as well as differential power processing (DPP) converters to handle the difference in power between PV submodules.
Maximum power point tracking (MPPT)
Alternating current photovoltaic (ACPV)
Microinverter
dc optimizer
Photovoltaic (PV)
Power Conversion
Power Electronics
Photovoltaic (PV) Submodule
Photovoltaic (PV) SubMIC
Module Integrated
Solar
Energy Harvest
Global maximum power point tracking (MPPT)
Local maximum power point tracking (MPPT)
Distributed maximum power point tracking (MPPT)
Thu, 16 Jan 2014 18:25:42 GMTCircuit level delay and power analysis of graphene nano-ribbon field-effect transistors using monte carlo simulations and standard cell library characterization
http://hdl.handle.net/2142/46893
Circuit level delay and power analysis of graphene nano-ribbon field-effect transistors using monte carlo simulations and standard cell library characterization
Sangai, Amit
Graphene nano-ribbon (GNR) transistors have emerged as a promising candidate to replace traditional silicon transistors in future scaled technologies. Since these devices are very small, the impact of process variation on the circuit’s performance is very large.
In this work, we study the impact of process variations on the delay and power of various types of circuits by considering the transistor-level and gate- level impact of the different technology. HSPICE based simulations are per- formed to study smaller circuits and the pros and cons of using GNR transistors. Monte Carlo simulations are performed to analyse the sensitivity of delay and power to the change in transistor parameters and also to analyse the situation when all parameters vary together during fabrication. Standard cell library design and characterisation is also explained, which is essential to perform simulations on large circuits that HSPICE is unable to handle. The standard cell library is also constructed and tested on four circuits to validate the library.
Graphene
Monte Carlo
Graphene nanoribbon field-effect transistor (GNRFET)
Graphene nano-ribbon
Thu, 16 Jan 2014 18:25:31 GMTFully automatic vision-based system for vehicle crash prediction and recognition
http://hdl.handle.net/2142/46867
Fully automatic vision-based system for vehicle crash prediction and recognition
Khorrami, Pooya
Just as they were half a century ago, automobile accidents are, unfortunately, one of the leading causes of death today. Therefore, it is no surprise that automated traffic analysis systems generate a supreme amount of interest. Despite the rapid advances in technology today, many traffic monitoring systems require substantial amounts of careful annotation. As such, a fully automated traffic analysis system that can perform accident prediction would be highly beneficial to multiple parties. Such systems would make a traffic analyst's workload more manageable and would provide a more sophisticated tool for determining the root causes of traffic accidents.
In this thesis, we present an automatic vision-based system for both accident prediction and recognition. Our method first detects and tracks vehicles using Robust Principal Component Analysis (Robust PCA) and Kalman Filters in order to extract trajectories. Pairs of vehicles trajectories are then segmented and classified by a Support Vector Machine (SVM) in order to determine the likelihood of a collision. We also tackle the problem of accident recognition by classifying crashing trajectory pairs into distinct categories. An ontology is used to define the relationships between the accident types and to train a tree-based classifier for recognition. We demonstrate the effectiveness of each algorithm by evaluating them on a crash dataset provided by the Toyota Motor Corp.
Vision-Based Surveillance System
Robust Principal Component Analysis (PCA)
Kalman Filter
Accident Prediction
Accident Recognition
Background Subtraction
Object Tracking
Thu, 16 Jan 2014 18:19:08 GMTStudies of some issues in solving surface integral equations and the equivalence principle algorithm
http://hdl.handle.net/2142/46863
Studies of some issues in solving surface integral equations and the equivalence principle algorithm
Sarker, Palash
This thesis describes the progressive development analysis of the equivalence principle
algorithm (EPA) beginning with the development of various operators that constitute
it. We begin with the formulation for the electric field integral equation (EFIE),
and visit the necessary treatment of the magnetic field integral equation (MFIE),
the combined field integral equation (CFIE), and the Poggio-Miller-Chu-Harrington-
Wu-Tsai (PMCHWT) along with various singularity extraction schemes before we
develop the EPA relations. The explicit expressions of the translation operators are
also derived. The EFIE, the MFIE, and the CFIE formulations are used to verify the
accuracy of such operators as the L, the K, the nX L, and the n XK operators that
are at the heart of the EPA formulation. Very detailed derivations and analyses of
these operators with proper scaling factors are included to avoid the inaccuracy due
to iterative solvers. Later we provide pertinent results for all of them for verification
and comparison.
Equivalence principle algorithm (EPA)
equivalence principle operator (EPO)
radar cross section (RCS)
translation operator (TO)
tap
vertical-vertical (VV)
Thu, 16 Jan 2014 18:19:02 GMTInformation trust, inference and transfer in social and information networks
http://hdl.handle.net/2142/46854
Information trust, inference and transfer in social and information networks
Qi, GuoJun
In this thesis, our overarching goal is to aggregate crowdsourced information that is collected from
computing systems based on social networks and represented in information networks. Due to the autonomous nature of
such a social computing paradigm, the crowdsourced information is often subject to low quality, contributed by susceptible
information sources without a reliant quality control scheme. Thus, to reveal the trustworthiness of the involved information sources, we aim to explore the social dependency behind the social networks where information contributors are prone to be influenced by each other. We explored the impact of such social dependency between sources on the information trust, aggregation and quality in social computing models. On the other hand, we will also investigate the structure underlying information shared by sources to reveal their trustworthiness.
Our study will deepen our understanding of the patterns and behaviors of information sources and their reliability from both social and information aspects. Several closely related problems are investigated in this thesis: (1) the source trustworthiness, which aims to distinguish the untrustworthy sources from the trustworthy ones; (2) social signal processing, which aims to aggregate the multi-source contributed information to recover the true signals behind the problems such as the correct answers to a question and the true labels for an image; (3) the social dependency, which reveals the mutual influences among different sources; and (4) the nature of information structure, such as the information dependency underlying low-rank structure and visual similarities. Our goal is to propose a unified probabilistic model to explain the social and information phenomena behind these problems. In this thesis, we designed several algorithms which are tested in several real social and information network scenarios. Superior performances have been achieved compared with many existing state-of-the-art technologies in the areas.
information trust
information inference
information transfer
information networks
social networks
Thu, 16 Jan 2014 18:18:46 GMTExploiting wireless broadcast property to improve performance of distributed algorithms and mac protocols in wireless networks
http://hdl.handle.net/2142/46850
Exploiting wireless broadcast property to improve performance of distributed algorithms and mac protocols in wireless networks
Hosseinabadi, Ghazale
Because a wireless channel is a shared medium, messages sent on the wireless links might be overheard by the neighboring stations. The information obtained from the overheard messages can be used in order to design more efficient distributed algorithms as well as MAC protocols for wireless networks. We exploit the wireless broadcast property in three different aspects.
First, we design mutual exclusion algorithms for wireless networks in which opportunistic packet overhearing is exploited to decrease the number of transmitted messages as well as the delay of the algorithm. Second, we design a distributed and dynamically adaptive MAC protocol for wireless networks, called Token-DCF. In Token-DCF an implicit token passing algorithm is proposed to reduce idle and collision times of the random access mechanism of
IEEE 802.11 DCF protocol. In Token-DCF, packet overhearing is employed
to exchange scheduling information across the network. Third, we consider a dense deployment of wireless LANs and we propose Concurrent-MAC, a MAC protocol for increasing concurrent transmissions in dense wireless LANs. In Concurrent-MAC, based on SINR values between stations and access points (APs), sets of concurrent transmitters are identified by the backhaul of APs. A station gaining access to the channel schedules a set of its neighbors for concurrent transmissions. Neighbors chosen for concurrent transmission can start transmitting on the channel immediately after they overhear the privilege given to them for concurrent transmission.
Wireless Networks
Media Access Control (MAC) Protocols
Distributed Algorithms
Thu, 16 Jan 2014 18:18:37 GMTStructured concept recycling by probabilistic logic ontology tree
http://hdl.handle.net/2142/46848
Structured concept recycling by probabilistic logic ontology tree
Chang, Shiyu
Recent advances in multimedia research have generated a large collection of concept models, e.g., LSCOM and Mediamill 101, which have become accessible to other researchers. While most current research efforts still focus
on building new concepts from scratch, little effort has been made to construct new concepts upon the existing models already in the "warehouse". To address this issue, we have developed a new framework in this thesis, termed LEarning structured model by probabilistic loGic Ontology (LEGO) to seamlessly integrate both the new target training examples and the existing
primitive concept models. LEGO treats the primitive concept models
as a Lego toy to potentially construct an unlimited vocabulary of new concepts. Specifically, LEGO first formulates the logic operations to be the Lego
connectors used to combine existing concept models hierarchically in probabilistic logic ontology trees. LEGO then simultaneously incorporates new target training information to efficiently disambiguate the underlying logic
tree and correct the error propagation. We present extensive experimental results on a large vehicle domain data set from ImageNet and demonstrate
significantly superior performance over existing state-of-the-art approaches which build new concept models from scratch.
Multimedia LEarning structured model by probabilistic loGic Ontology (LEGO)
Concept recycling
Model warehouse
Probabilistic logic ontology tree
Logical operations
Thu, 16 Jan 2014 18:18:31 GMT