Electric power systems are huge real time energy distribution networks where accurate short term load forecasting (STLF) plays an essential role. This thesis is an effort to comprehensively investigate new and advanced neural network (NN) architectures to perform STLF. Two hybrid and two 3-layered NN architectures are introduced. Each network is individually tested to generate weekday and weekend forecasts using data from three jurisdictions of Canada. Overall findings suggest that 3-layered cascaded NN have outperformed almost all others for weekday forecasts. For weekend forecasts 3-layered feed forward NN produced most accurate results. Recurrent and hybrid networks performed well during peak hours but due to occurrence of constant high error spikes were not able to achieve high accuracy.

Made available in DSpace on 2014-06-12T15:58:19Z (GMT). No. of bitstreams: 2 arquivo3464_1.pdf: 2672787 bytes, checksum: 9bee33c2153182c2ce64b9027453243a (MD5) license.txt: 1748 bytes, checksum: 8a4605be74aa9ea9d79846c1fba20a33 (MD5) Previous issue date: 2011 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Sistemas computacionais estão em constante evolução para satisfazer crescimentos na demanda, ou novas exigências dos usuários. A administração desses sistemas requer decisões que sejam capazes de prover o nível mais alto nas métricas de desempenho e dependabilidade, com mudanças mínimas `a configuração existente. É comum realizar análises de desempenho, confiabilidade, disponibilidade e performabilidade de sistemas através de modelos analíticos, e as cadeias de Markov representam um dos formalismos matemáticos mais utilizados, permitindo estimar algumas métricas de interesse, dado um conjunto de parâmetros de entrada. No entanto, a análise de sensibilidade, quando feita, é executada simplesmente variando o conjunto de parâmetros dentro de suas faixas de valores e resolvendo repetidamente o modelo escolhido. A análise de sensibilidade diferencial permite a quem está modelando encontrar gargalos de uma maneira mais sistemática e eficiente. Este trabalho apresenta uma abordagem automatizada para análise de sensibilidade, e almeja guiar a melhoria de sistemas computacionais. A abordagem proposta é capaz de acelerar o processo de tomada de decisão, no que se refere a optimização de ajustes de hardware e software, além da aquisição e substituição de componentes. Tal metodologia usa as cadeias de Markov como técnica de modelagem formal, e a análise de sensibilidade desses modelos, preenchendo algumas lacunas encontradas na literatura sobre análise de sensibilidade. Por fim, a análise de sensibilidade de alguns sistemas distribuídos selecionados, conduzida neste trabalho, destaca gargalos nestes sistemas e fornece exemplos da acurácia da metodologia proposta, assim como ilustra sua aplicabilidade

In multi-level decision making (DM) approaches, the final decision is reached by going through a finite number of DM levels. Usually, in each level, a raw decision is produced first and then a suitable decision fusion technique is employed to merge the lower level decisions with the raw decision in the construction of the final decision of the present level. The basic difficulty in these approaches is the determination of how the consecutive levels should interact with each other. In this thesis, two different multi-level DM models have been proposed. The main idea in the first model, &ldquo / hierarchical DM&rdquo / (HDM), is to transfer the decisions of previous hierarchical levels to an upper hierarchy with some reliability values. These decisions are then fused using a suitable decision fusion technique to attain more consistent decisions at an upper level. The second model &ldquo / local DM in multiplelevels&rdquo / (LDM-ML) depends on what may be called as local DM process. Instead of designing an agent to perform globally, designing relatively simple agents which are supposed to work in local regions is the essence of the second idea. Final decision is partially constructed by contribution of a sufficient number of local DM agents. A successful local agent is retained in the agent pool whereas a local agent not successful enough is eliminated and removed from the agent pool. These models have been applied on two case studies associated with fault detection in a four-tank system and prediction of lotto sales.

Electric energy constitutes one of the most crucial elements to almost every aspect of life of people. The modern electric power systems face several challenges such as efficiency, economics, sustainability, and reliability. Increase in electrical energy demand, distributed generations, integration of uncertain renewable energy resources, and demand side management are among the main underlying reasons of such growing complexity. Additionally, the elements of power systems are often vulnerable to failures because of many reasons, such as system limits, weak conditions, unexpected events, hidden failures, human errors, terrorist attacks, and natural disasters. One common factor complicating the operation of electrical power systems is the underlying uncertainties from the demands, supplies and failures of system components. Stochastic programming provides a mathematical framework for decision making under uncertainty. It enables a decision maker to incorporate some knowledge of the intrinsic uncertainty into the decision making process. In this dissertation, we focus on application of two-stage and multistage stochastic programming approaches to electric energy systems modeling and optimization. Particularly, we develop models and algorithms addressing the sustainability and reliability issues in power systems. First, we consider how to improve the reliability of power systems under severe failures or contingencies prone to cascading blackouts by so called islanding operations. We present a two-stage stochastic mixed-integer model to find optimal islanding operations as a powerful preventive action against cascading failures in case of extreme contingencies. Further, we study the properties of this problem and propose efficient solution methods to solve this problem for large-scale power systems. We present the numerical results showing the effectiveness of the model and investigate the performance of the solution methods. Next, we address the sustainability issue considering the integration of renewable energy resources into production planning of energy-intensive manufacturing industries. Recently, a growing number of manufacturing companies are considering renewable energies to meet their energy requirements to move towards green manufacturing as well as decreasing their energy costs. However, the intermittent nature of renewable energies imposes several difficulties in long term planning of how to efficiently exploit renewables. In this study, we propose a scheme for manufacturing companies to use onsite and grid renewable energies provided by their own investments and energy utilities as well as conventional grid energy to satisfy their energy requirements. We propose a multistage stochastic programming model and study an efficient solution method to solve this problem. We examine the proposed framework on a test case simulated based on a real-world semiconductor company. Moreover, we evaluate long-term profitability of such scheme via so called value of multistage stochastic programming.

System models have been used to improve water management and environmental decision making. In spite of the many existing mathematical models and tools that attempt to improve environmental decision making, few efforts have been made to identify how scarce resources (e.g., water, budget) can be more efficiently allocated to improve the environmental and ecological performance of different ecosystems (e.g., wetland habitat). This dissertation presents a set of management tools to improve the environmental and ecological performance. These tools are described in three studies. First, a simple optimization model is developed to help regulators and watershed managers determine cost-effective best management practices (BMPs) to reduce phosphorus load at the Echo Reservoir Watershed, Utah. The model minimizes the costs of BMP implementation to achieve a specified phosphorus load reduction target. Second, a novel approach is developed to quantify wetland habitat performance. This performance metric is embedded in a new optimization model to recommend water allocations and invasive vegetation control in wetlands. Model recommendations are subject to constraints such as water availability, spatial connectivity of wetland, hydraulic infrastructure capacities, vegetation growth and responses to management, plus financial and time resources available to allocate water and invasive vegetation control. Third, an agent-based model is developed to simulate the spread of the invasive Phragmites australis (common reed), one of the most successful invasive plant species in wetlands. Results of the agent-based model are embedded into an optimization model (developed in the second study) to recommend invasive vegetation control actions. The second and third studies were applied at the Bear River Migratory Bird Refuge, which is the largest wetland complex on the Great Salt Lake, Utah. These three studies provide a set of decision-support tools that recommend: (1) BMPs to reduce phosphorus loading in a watershed, (2) management strategies to improve wetland bird habitat, and (3) control strategies to minimize invasive Phragmites spread. Together, these models provide important insights and recommendations for managers to make informed decisions to manage excess nutrients in water bodies as well as to improve wetland management.

No description available.

Photovoltaic technology is among the best tools our civilization has to reduce the emissions of greenhouse gas that are currently altering the atmosphere composition of our planet. The idea of using photovoltaic surfaces on the envelope of buildings is called with the acronym of BIPV (building integrated photovoltaics), it offers the advantage of producing energy in the same location of the demand for electricity. Furthermore, BIPV allows to save monetary and environmental costs by substituting building materials with photovoltaic collectors. As every technology,BIPV follows an adoption pattern that is bringing it from a very limited niche product to a pervasive one. Nevertheless, the adoption rate of BIPV appears to be slow, and the industry has offered little opportunities of business for its stakeholders over the last 20 years. There are multiple reasons for this sluggish growth, and a considerable body of scientific literature has offered potential solutions to the problem. The building industry is notoriously slow in picking up innovation, furthermore the BIPV material needs to compete with much more mature, versatile and often cheaper cladding technologies and materials. Numerous research endeavors are focusing on the development of new BIPV claddings to have diversified colors, dimensions, shapes and other properties. The argument is that the technology is not mature and thus cannot be adopted by the bulk of architects and designers. Unfortunately, the premium characteristics of these new materials often come with a higher price and a reduced efficiency, thus reducing their market potential. Other research endeavors, among which this thesis, are focusing on the design of buildings: trying to include the use of photovoltaics into the architectural practice through education and software development. Numerous software has been developed over the last 20 years with the aim of calculating the productivity or the economic outlook of a BIPV system. The main difference between the existing software and the method presented here lies in the following fact: previously, the capacity and positions of a BIPV system are required as input for the calculation of performance, in this method the capacity and positions of the BIPV system are given as the output of an optimization process. A designer whois skeptical or disengaged about the use of BIPV could be induced to avoid its use entirely by the discouraging simulation results given by the lack of a techno-economic optimal configuration. Conversely, a designer who opt for a premium architectural PV material would, thank to the methodology shown, be able to assess the impact its unitary cost has on the optimal BIPV capacity of the building. Ultimately, the method presented provides new knowledge to the designer regarding the use of BIPV on his building, hopefully this can facilitate the spread of BIPV technology. The method described was translated into a software tool to find the best positions and number of PV surfaces over the envelope of the building and the best associated battery capacity. The tool is based on the combined use of ray-tracing (for irradiation calculation) and optimization algorithms, its use led to the following conclusions: • BIPV is profitable under a wide range of assumptions if installedin the correct capacities • 20% of the residential electric demand can easily be covered by PV without the need for electric storage and in a profitable way • Despite an interesting rate of return of the investment, the payback time was generally found to be long (over 10 years) • More research is needed to assess the risk on the investment on BIPV: if found to be low, future financial mechanisms could increase its spread despite the long payback time • The optimal capacity in energy terms (i.e. the energy consumed on-site minus the energy used to produce a BIPV system) tends to be far higher than any techno-economic optimum • The specific equivalent CO2 emissions for an NPV optimal system have been found to be between 70 and 123 [kg CO2 eq/MWh] under the range of assumptions applied • The installation of optimal BIPV capacity could change the overall residential CO2 emission of -12%, +13%, -29% in England, France and Greece respectively • despite the non optimal placement of a BIPV system compared to a ground mounted, south oriented one, and despite the noncontemporaneity of production and consumption, the BIPV still easily outperforms the energy mix of most countries when optimized for maximum NPV. • The part of the building envelope that have the most annual irradiation (i.e. the roof) should not necessarily host the entirety of the system as other facades might have an advantage in terms of matching production and consumption times. • when different scenarios are made in terms of techno-economic input parameters (e.g. degradation of the system, future costs of maintenance, future variation of electricity price etc..) larger capacities are optimal for optimistic outlooks and vice-versa • the optimal capacity for the expected scenario (i.e. the 50 % ile) can be considered robust as it performs close to the optimum in optimistic and pessimistic scenarios alike. • a reduction in price for the electric storage appears to have a positive effect on the optimal capacity of PV installed for the case study considered. • when a group of households is optimized separately V.S. aggregated together, the aggregation have a huge positive effect on all KPIs of the resulting system: in the NPV optimal system of a case study examined the installed capacity ( +118%), the NPV ( +262.2%) and the self-sufficiency( +51%) improved thanks to aggregation.

A geração de grades horárias de qualidade é um fator crítico em qualquer instituição de ensino, tanto em escolas de ensino fundamental/médio como em universidades. Este problema é considerado complexo, pois devem ser relacionados e otimizados diversos recursos, tais como horários, disciplinas, professores e alunos. Em grande parte das instituições de ensino, a geração de grades horárias é realizada manualmente, o que vem a tornar este processo custoso e sujeito a falhas. Diversas abordagens são também encontradas na literatura para resolução deste problema, nas quais foram aplicados métodos de busca estocástica, devido à sua inerente complexidade. As estratégias de busca formuladas e comparadas no presente trabalho foram baseadas no uso de algoritmos genéticos e de sistemas imunológicos artificiais. Tais técnicas foram capazes de fornecer soluções de qualidade para o problema de geração automática de grades horárias. Neste trabalho foram desenvolvidos dois sistemas de apoio à decisão, nos quais foram combinadas técnicas heurísticas aos algoritmos genéticos e ao algoritmo de seleção clonal. O propósito desta investigação é realizar uma análise comparativa entre as duas técnicas a fim de verificar qual delas apresenta resultados mais promissores para a resolução do problema de geração automática de grades horárias. / The generation of timetables with good quality is a critical factor in any educational institution. This is considered a complex problem because it involves several types of information, such as schedules, course subjects, teachers and students. Several search strategies have been applied to solve timetabling problems, whose constraints may vary from one educational institution to another. Most educational institutions still prepare their timetables manually, which is a highly time-consuming process and subjected to errors. Several approaches to solve this problem are also found in technical studies, which use stochastic search methods due to the problems complexity. The search optimization methods used in this work to solve the timetabling problem are genetic algorithms and the clonal selection algorithm, whose satisfactory results when applied to optimization problems are reported in the literature. Two decision support systems were developed in this work, combining heuristic techniques with the genetic algorithms and the clonal selection algorithm. The purpose of this research is to make a comparative analysis of these two techniques in order to determine which one offers the most promising results for solving the timetabling problem.

A geração de grades horárias de qualidade é um fator crítico em qualquer instituição de ensino, tanto em escolas de ensino fundamental/médio como em universidades. Este problema é considerado complexo, pois devem ser relacionados e otimizados diversos recursos, tais como horários, disciplinas, professores e alunos. Em grande parte das instituições de ensino, a geração de grades horárias é realizada manualmente, o que vem a tornar este processo custoso e sujeito a falhas. Diversas abordagens são também encontradas na literatura para resolução deste problema, nas quais foram aplicados métodos de busca estocástica, devido à sua inerente complexidade. As estratégias de busca formuladas e comparadas no presente trabalho foram baseadas no uso de algoritmos genéticos e de sistemas imunológicos artificiais. Tais técnicas foram capazes de fornecer soluções de qualidade para o problema de geração automática de grades horárias. Neste trabalho foram desenvolvidos dois sistemas de apoio à decisão, nos quais foram combinadas técnicas heurísticas aos algoritmos genéticos e ao algoritmo de seleção clonal. O propósito desta investigação é realizar uma análise comparativa entre as duas técnicas a fim de verificar qual delas apresenta resultados mais promissores para a resolução do problema de geração automática de grades horárias. / The generation of timetables with good quality is a critical factor in any educational institution. This is considered a complex problem because it involves several types of information, such as schedules, course subjects, teachers and students. Several search strategies have been applied to solve timetabling problems, whose constraints may vary from one educational institution to another. Most educational institutions still prepare their timetables manually, which is a highly time-consuming process and subjected to errors. Several approaches to solve this problem are also found in technical studies, which use stochastic search methods due to the problems complexity. The search optimization methods used in this work to solve the timetabling problem are genetic algorithms and the clonal selection algorithm, whose satisfactory results when applied to optimization problems are reported in the literature. Two decision support systems were developed in this work, combining heuristic techniques with the genetic algorithms and the clonal selection algorithm. The purpose of this research is to make a comparative analysis of these two techniques in order to determine which one offers the most promising results for solving the timetabling problem.

No description available.