In recent years, there have been significant changes in power systems due to the integration of renewables, distributed generation, switched power loads, and energy storage systems, etc. Locally these AC/DC microgrids include both DC generation (such as solar PV) and AC generation (such as wind generation), various DC and AC loads, converters and inverters, and energy storage systems, such as storage batteries and supercapacitors. DC systems are often characterized as low inertia systems whereas AC generation and systems are usually high inertia and high time constant systems. As such, various components of the microgrid will have different temporal characteristics in case of disturbances, such as short circuit, load switchings, etc. which may lead to instability of the microgrid. This research develops the first principle model for coupling the AC and the DC subsystem of an integrated AC/DC microgrid utilizing the dq-framework. The developed model is highly nonlinear and captures the dynamic interaction between the AC and DC subsystems of the microgrid. Lyapunov stability is used to evaluate the stability of the complete system. Simulation results show that the AC and DC subsystems are tightly dynamically coupled so that any disturbance in one subsystem induces transients in the other subsystem. Induced transients due to pulse loads on the AC and DC subsystems clearly show that generator damper winding alone may not be enough to mitigate transients in the microgrid. Addition of prime mover and excitation system controllers for the generator improves the transients primarily on the AC subsystem. Thus, a battery storage with a charge/discharge controller was also added to the DC subsystem. Simulations of the AC/DC microgrid with all three controllers validate the smooth operation of the system for all types of disturbances. The proposed method can be extended in modeling microgrid with multiple generators and various types of loads. / Electrical and Computer Engineering

This dissertation focuses on the modeling and control of WECS (Wind Energy Conversion System) in a multi-machine system. As one of the fastest growing renewable energy resources, the trend of wind energy changes to variable speed wind turbines. The concept of the variable speed is based on the variable speed according to the instantaneous wind speed of wind turbines. Since the utility grid requires the stable frequency and magnitude voltages, there must be grid interconnection of the wind farm and the utility grid. The grid interconnection must support the concept of the variable speed wind turbines. Since each wind turbine locates in a different location in a wind site, it receives the different wind speed. Hence the grid interconnection must convert the variable frequency and magnitude output voltages of the wind turbines to a synchronous frequency and magnitude voltages associated to the grid. With the new technologies of power semiconductor devices, the power converter can operate with high voltage, high current, and high switching frequency. This results in a higher power capacity of a wind farm. Nonetheless, the power converters generate harmonic distortions to the utility grid. The harmonic distortions components in the voltages and currents of the grid degrade the power quality. This results in the damage of electrical components in the power system such as capacitor banks, inductors, protection devices, etc. The harmonic distortions can be reduced with the technology of the multi-level inverter. It is required that the wind energy provides the real and reactive power control for frequency and voltage stability. In order to achieve the power control, the modeling and control of the power electronic grid interconnection is presented in this dissertation. The grid interconnection is modeled with linearization techniques. The models in frequency domain in the form of transfer functions are used to design the compensators in the control system. The model is considered as a SISO (Single Input Single Output) system to design the compensators in SISO tool of MATLAB. The selected control system is current control that can control the real and reactive powers independently. Furthermore, since the grid interconnection is modeled separately for each sub-system, the control system is verified with integration of the sub-systems. The grid interconnection is modeled in Simulink and simulated in the PSCAD. In reality, the power system is comprised of multi-machines. They affect the power system stability, reliability, and quality. The dynamic modeling of an aggregated wind farm with synchronous generator and grid interconnection in a multi-machine system is presented. The test system is a 10-bus system with three generators and three loads. The dynamic modeling involves the power flow calculations that determine the equilibrium points of the system. The system is modeled with differential equations of wind turbines, synchronous generators, and grid interconnection. The system is modeled in the time domain in state space form. The system characteristics can be determined by poles or eigen values obtained from the characteristic equations. Since the system is MIMO (Multi Input Multi Output) system, the optimal control theory is used to reduce the deviation of system behaviors during disturbances. The LQR (Linear Quadratic Regulator) is utilized to control the system with eigen value assignment method. Simulation results in Simulink are illustrated. / Ph. D.

This dissertation explores the design, and implementation of articulated robotic tail mechanisms onboard reduced degree-of-freedom (DOF) legged robots to augment performance in terms of stability and maneuverability. Fundamentally, this research is motivated by the question of how to improve the stability and maneuverability of legged robots. The conventional approach to address these challenges is to utilize leg mechanisms that are composed of three or more active DOFs that are controlled simultaneously to provide propulsion, maneuvering, and stabilization. However, animals such as lizards and cheetahs have been observed to utilize their tails to aid in these functionalities. It is hypothesized that by using an articulated tail mechanism to aid in these functionalities onboard a legged robot, the burden on the robot's legs to simultaneously maneuver and stabilize the robot may be reduced. This could allow for simplification of the leg's design and control algorithms. In recent years, significant progress has been accomplished in the field of robotic tail implementation onboard mobile robots. However, the main limitation of this work stems from the proposed tail designs, the majority of which are composed of rigid single-body pendulums that provide a constrained workspace for center-of-mass positioning, an important characteristics for inertial adjustment applications. Inspired by lizards and cheetahs that adjust their body orientation using flexible tail motions, two novel articulated, cable driven, serpentine-like tail mechanisms are proposed. The first is the Roll-Revolute-Revolute Tail which is a 3-DOF mechanism, designed for implementation onboard a quadruped robot, that is capable of forming two mechanically decoupled tail curvatures via an s-shaped cable routing scheme and gear train system. The second is a the Discrete Modular Serpentine Tail, designed for implementation onboard a biped robot, which is a modular two-DOF mechanism that distributes motion amongst links via a multi-diameter pulley. Both tail designs utilize a cable transmission system where cables are routed about circular contoured links that maintain equal antagonistic cable displacements that can produce controlled articulated tail curvatures using a single active-DOF. Furthermore, analysis and experimental results have been presented to demonstrate the effectiveness of an articulated tail's ability to: 1) increase the manifold for center-of-mass positioning, and 2) generate enhanced inertial loading relative to conventionally implemented pendulum-like tails. In order to test the tails ability to augment the performance of legged robots, a novel Robotic Modular Leg (RML) is proposed to construct both a reduced-DOF quadrupedal and bipedal experimental platform. The RML is a modular two-DOF leg mechanism composed of two serially connected four-bar mechanisms that utilizes kinematic constraints to maintain a parallel orientation between it's flat foot and body without the use of an actuated ankle. A passive suspension system integrated into the foot enables the dissipation of impact energy and maintains a stable four point-of-contact support polygon on both flat and uneven terrain. Modeling of the combined legged robotic systems and attached articulated tails has led to the derivation of dynamic formulations that were analyzed to scale articulated tails onboard legged robots to maximize inertial adjustment capabilities resulting from tail motions and design a control scheme for tail-aided maneuvering. The tail prototypes, in conjunction with virtual simulations of the quadruped and biped robot, were used in experiments and simulations to implement and analyze the methods for maneuvering and stabilizing the proposed legged robots. Results successfully demonstrate the tails' ability to augment the performance of reduced-DOF legged robots by enabling comparable walking criteria with respect to conventional legged robots. This research provides a firm foundation for future work involving design and implementation of articulated tails onboard legged robots for enhanced inertial adjustment applications. / Ph. D.

Geologic sequestration has been suggested as a viable method for greenhouse gas emission reduction. Regional studies of CO₂ storage capacity are used to estimate available storage, yet little work has been done to tie site specific results to regional estimates. In this study, a 9,258,880 acre (37469.4 km²) area of the coastal and offshore Texas Miocene interval is evaluated for CO₂ storage capacity using a static volumetric approach, which is essentially a discounted pore volume calculation. Capacity is calculated for the Miocene interval above overpressure depth and below depths where CO₂ is not supercritical. The goal of this study is to determine the effectiveness of such a regional capacity assessment, by performing refinement techniques that include simple analytical and complex reservoir injection simulations. Initial refinement of regional estimates is performed through net sand picking which is used instead of the gross thickness assumed in the standard regional calculation. The efficiency factor is recalculated to exclude net-to-gross considerations, and a net storage capacity estimate is calculated. Initial reservoir-scale refinement is performed by simulating injection into a seismically mapped saline reservoir, near San Luis Pass. The refinement uses a simplified analytical solution that solves for pressure and fluid front evolution through time (Jain and Bryant, 2011). Porosity, permeability, and irreducible water saturation are varied to generate model runs for 6,206 samples populated using data from the Atlas of Northern Gulf of Mexico Gas and Oil Reservoirs (Seni, 2006). As a final refinement step, a 3D dynamic model mesh is generated. Nine model cases are generated for homogeneous, statistically heterogeneous, and seismic-based heterogeneous meshes to observe the effect of various geologic parameters on injection capacity. We observe downward revisions (decreases) in total capacity estimation with increasingly refined geologic data and scale. Results show that estimates of storage capacity can decrease significantly (by as much as 88%) for the single geologic setting investigated. Though this decrease depends on the criteria used for capacity comparison and varies within a given region, it serves to illustrate the potential overestimation of regional capacity assessments compared to estimates that include additional geologic complexity at the reservoir scale. / text

<p>In this work, speaker characteristic modeling has been applied in the fields of automatic speech recognition (ASR) and automatic speaker verification (ASV). In ASR, a key problem is that acoustic mismatch between training and test conditions degrade classification per- formance. In this work, a child exemplifies a speaker not represented in training data and methods to reduce the spectral mismatch are devised and evaluated. To reduce the acoustic mismatch, predictive modeling based on spectral speech transformation is applied. Follow- ing this approach, a model suitable for a target speaker, not well represented in the training data, is estimated and synthesized by applying vocal tract predictive modeling (VTPM). In this thesis, the traditional static modeling on the utterance level is extended to dynamic modeling. This is accomplished by operating also on sub-utterance units, such as phonemes, phone-realizations, sub-phone realizations and sound frames.</p><p>Initial experiments shows that adaptation of an acoustic model trained on adult speech significantly reduced the word error rate of ASR for children, but not to the level of a model trained on children’s speech. Multi-speaker-group training provided an acoustic model that performed recognition for both adults and children within the same model at almost the same accuracy as speaker-group dedicated models, with no added model complexity. In the analysis of the cause of errors, body height of the child was shown to be correlated to word error rate.</p><p>A further result is that the computationally demanding iterative recognition process in standard VTLN can be replaced by synthetically extending the vocal tract length distribution in the training data. A multi-warp model is trained on the extended data and recognition is performed in a single pass. The accuracy is similar to that of the standard technique.</p><p>A concluding experiment in ASR shows that the word error rate can be reduced by ex- tending a static vocal tract length compensation parameter into a temporal parameter track. A key component to reach this improvement was provided by a novel joint two-level opti- mization process. In the process, the track was determined as a composition of a static and a dynamic component, which were simultaneously optimized on the utterance and sub- utterance level respectively. This had the principal advantage of limiting the modulation am- plitude of the track to what is realistic for an individual speaker. The recognition error rate was reduced by 10% relative compared with that of a standard utterance-specific estimation technique.</p><p>The techniques devised and evaluated can also be applied to other speaker characteristic properties, which exhibit a dynamic nature.</p><p>An excursion into ASV led to the proposal of a statistical speaker population model. The model represents an alternative approach for determining the reject/accept threshold in an ASV system instead of the commonly used direct estimation on a set of client and impos- tor utterances. This is especially valuable in applications where a low false reject or false ac- cept rate is required. In these cases, the number of errors is often too few to estimate a reli- able threshold using the direct method. The results are encouraging but need to be verified on a larger database.</p> / Pf-Star / KOBRA

<p>Bayesian methods offer a flexible and convenient probabilistic learning framework to extract interpretable knowledge from complex and structured data. Such methods can characterize dependencies among multiple levels of hidden variables and share statistical strength across heterogeneous sources. In the first part of this dissertation, we develop two dependent variational inference methods for full posterior approximation in non-conjugate Bayesian models through hierarchical mixture- and copula-based variational proposals, respectively. The proposed methods move beyond the widely used factorized approximation to the posterior and provide generic applicability to a broad class of probabilistic models with minimal model-specific derivations. In the second part of this dissertation, we design probabilistic graphical models to accommodate multimodal data, describe dynamical behaviors and account for task heterogeneity. In particular, the sparse latent factor model is able to reveal common low-dimensional structures from high-dimensional data. We demonstrate the effectiveness of the proposed statistical learning methods on both synthetic and real-world data.</p> / Dissertation

Este trabalho trata da modelagem dinâmica e do projeto de sistemas de controle para um robô manipulador de arquitetura paralela assimétrica de três graus de liberdade, correspondente a três translações de seu efetuador no espaço tridimensional, concebido para tarefas de pega-e-põe (pick-and-place). Dentre os desenvolvimentos teóricos, procurou-se estender, para toda gama de robôs paralelos e topologicamente assimétricos, a abrangência dos procedimentos aplicados inicialmente a este caso específico. Foram empregados o Método de Lagrange e o Princípio dos Trabalhos Virtuais na obtenção de modelos dinâmicos baseados em hipóteses simplificadoras de corpos rígidos de massas concentradas e de massas distribuídas. Para o projeto de controladores, foram utilizadas as técnicas de torque computado e torque computado estendido. As formulações correspondentes a estas técnicas foram aprimoradas de forma a permitir o cálculo de esforços de controle de modo equivalente tanto nas coordenadas dos atuadores como nas coordenadas do efetuador e em conformidade com os requisitos de resposta dinâmica definidos para o robô. Tais requisitos podem ser inclusive anisotrópicos, o que se julga ser mais apropriado para robôs paralelos assimétricos. Particularmente em relação ao robô analisado, foram avaliados efeitos de simplificação do modelo dinâmico e da discretização do controlador (incluindo discretização com dupla frequência de amostragem) sobre os erros de controle de posição. Percebeu-se que os requisitos de alto desempenho das tarefas de pega-epõe levam o projeto do controlador ao limite de validade da hipótese simplificadora de corpo rígido e da capacidade de processamento do hardware de controle. / This work deals with the dynamic modeling and the design of control systems for an asymmetric parallel manipulator robot with three degrees of freedom, related to three translations of its end-effector in the tridimensional space, conceived for pick-andplace tasks. Among the theoretical developments, one of them was the extension of the proposed procedures, initially applied to this specific case, to the whole class of topologically asymmetric and parallel mechanisms. The Lagrange Method and the Principle of Virtual work were employed in order to build dynamic models based on rigid body assumption with either lumped or distributed mass simplification hypothesis. For the controllers design, computed torque and extended computed torque techniques were used. The formulations of such techniques were elaborated in order to allow the calculation of the control efforts to be executed equivalently both on the actuators coordinates and on the end-effector coordinates, in accordance with the dynamic response requirements defined for the robot. These requirements may be even anisotropic, what is considered to be more suitable for asymmetric parallel robots. Particularly about the analyzed robot, the effects of dynamic model simplification and controller discretization (including double sample rate discretization) on the position control errors were evaluated. It was realized that the high performance requirements for pick-and-place tasks push the controller design to the limit of validity of the rigid body assumption and of the control hardware processing capacity.

A dinâmica das paisagens e as consequentes alterações no uso e cobertura da terra podem ser analisadas a partir de modelagens espaciais, que auxiliam na interpretação das variáveis responsáveis pelas alterações na paisagem, bem como viabilizam a criação de cenários distintos visando ao conhecimento do impacto produzido pela inserção ou supressão de variáveis na simulação de mudanças da paisagem. Em Guarulhos, a instalação de grandes rodovias conectando o município de São Paulo ao Vale do Paraíba e, posteriormente, de um terminal aeroportuário, em consonância com diferentes períodos de crescimento econômico e expansão urbana, impactaram fortemente na formação da paisagem encontrada neste município, principalmente com a instalação dos grandes pátios industriais e bairros residenciais de média e baixa renda. Com isso, a presente pesquisa buscou avaliar as mudanças no uso e cobertura da terra a partir de produtos de sensoriamento remoto para identificar a influência advinda da instalação dos grandes pontos de conexão do município, por meio de modelagem com autômatos celulares (cellular automata - CA). Dentre os principais resultados obtidos está a influência positiva nas transições para a classe residencial regular exercida pela instalação do aeroporto no primeiro período de análise (1986 a 1993), e a repelência a essa transição nos anos seguintes. Resultados similares foram obtidos na avaliação da influência das grandes vias de acesso na implantação dos parques industriais e galpões logísticos. A modelagem dinâmica, por meio do software Dinamica-EGO, mostrou-se uma ferramenta bastante versátil para avaliação da influência das variáveis espaciais na dinâmica das paisagens urbanas. / The dynamics of landscapes and the consequent changes in land use and land cover can be analyzed according to spatial models, which helps the interpretation of the variables responsible for territorial changes, as well as enables the creation of distinct scenarios in order to know the impact produced by the insertion or suppression of variables in the simulation of landscape changes. In the municipal area of Guarulhos, the installation of large highways linking the municipality of São Paulo to the Paraíba Valley and an airport terminal afterwards, in accordance with different periods of economic growth and urban expansion, had a strong impact on the landscape found in this municipality, mostly because of the installation of large industrial yards and residential neighborhoods of medium and low income. Hence, this research aimed to evaluate the changes in land use and land cover according to remote sensing products, and it also aimed to try to identify the influence of the installation of big connection points of the municipality through modeling with cellular automata (CA). Among the main results obtained are the positive influence on the transitions to the regular residential class exercised by the airport installation in the first period of analysis (1986 to 1993), and the repellency of this transition in the following years. A similar result was obtained in the evaluation of the influence of the great access roads in the implantation of industrial parks and logistic sheds. Thus, the dynamic modeling through the Dynamica-EGO software proved to be a very versatile tool to evaluate the influence of spatial variables on the dynamics of urban landscapes.

"The goal of this project was to investigate the dynamic effects of incipient separation of industrial cam-follower systems. Typical industrial cam-follower systems include a force closed cam joint and a follower train containing both substantial mass and stiffness. Providing the cam and follower remain in contact, this is a one degree-of-freedom (DOF) system. It becomes a two-DOF system once the cam and follower separate or jump, creating two new natural frequencies, which bracket the original. The dynamic performance of the system as it passed through the lower of the two post-separation modes while on the verge of jump was investigated. A study was conducted to determine whether imperfections in the cam surface, while the contact force is on the brink of incipient separation, may cause a spontaneous switch to the two-DOF mode and begin vibration at resonance. A force-closed translating cam-follower train was designed for the investigation. The fixture is a physical realization of the two-mass mathematical model. Pro/Engineer was used to design the follower train, Mathcad and TK Solver were used to analyze the linkage and DYNACAM & Mathcad were used to dynamically model the system. The system is designed to be on the cusp of incipient separation when run. Experiments were carried out by bringing the system up to jump speed and then backing off the preload to get the system on the cusp of separation. Data were collected at the prejump, slight jump, and violently jumping stages. The time traces show the acceleration amplitudes grow to large peaks when the system is jumping. The frequency spectrum shows the two new natural frequencies growing in amplitude from non-existant in the prejump stage, to higher values in the violently jumping stage. The peak amplitudes of the phenomenon are small in magnitude compared to the harmonic content of the cam. It is concluded that the contribution of the two-DOF system natural frequencies is not a significant factor from a practical aspect. Although the actual jump phenomenon is of concern in high-speed applications, calculations show that if the follower system is designed sufficiently stiff then the two-DOF situation will not occur."

Avec l’accroissement considérable de la puissance de calcul, les simulations aux grandes échelles (SGE) sont maintenant utilisées de façon routinière dans de nombreuses applications d’ingénierie. Les modèles de combustion usuels utilisés dans les SGE sont le plus souvent basés sur une hypothèse d’équilibre entre le mouvement des structures turbulentes et le plissement de la surface de la flamme. Ils s’écrivent alors sous forme d’expressions algébriques fonctions de grandeurs connues aux échelles résolues ainsi que de paramètres dont l’ajustement est à la charge de l’utilisateur selon la configuration étudiée et les conditions opératoires. Le modèle dynamique récemment développé ajuste automatiquement au cours du calcul les paramètres de modélisation qui peuvent alors dépendre du temps et de l’espace. Cette thèse présente une étude détaillée d’un modèle dynamique pour la simulation aux grandes échelles de la combustion turbulente prémélangée. L’objectif est de caractériser, explorer les avantages et les inconvénients, appliquer et valider le modèle dynamique dans plusieurs configurations. / Large eddy simulation (LES) is currently applied in a wide range of engineering applications. Classical LES combustion models are based on algebraic expressions and assume equilibrium between turbulence and flame wrinkling which is generally not verified in many circumstances as the flame is laminar at early stages and progressively wrinkled by turbulent motions. In practice, this conceptual drawback has a strong consequence: every computation needs its own set of constants, i.e. any small change in the operating conditions or in the geometry requires an adjustment of model parameters. The dynamic model recently developed adjust automatically the flame wrinkling factor from the knowledge of resolved scales. Widely used to describe the unresolved turbulent transport, the dynamic approach remains underexplored in combustion despite its interesting potential. This thesis presents a detailed study of a dynamic wrinkling factor model for large eddy simulation of turbulent premixed combustion. The goal of this thesis is to characterize, unveil pros and cons, apply and validate the dynamic modeling in different flow configurations.