| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 5.8 MB | Adobe PDF |
Authors
Abstract(s)
O interesse em explorar a energia fotovoltaica tem crescido muito nos últimos anos. Tal
deve - se ao facto de ser um tipo de energia renovável muito disponível (visto que provém do
sol) e com um grande potencial de expansão e desenvolvimento. Nesse sentido, é essencial o
desenvolvimento de métodos que permitam prever e analisar o comportamento dos sistemas
fotovoltaicos, i.e., métodos que estimem com precisão os parâmetros dos módulos PV, sob
quaisquer condições de operação.
O objetivo principal desta dissertação é descrever com exatidão a modelação de sistemas
fotovoltaicos. Primeiramente é feita uma apresentação sobre os tipos de tecnologias
fotovoltaicas existentes atualmente e também sobre o funcionamento dos sistemas
fotovoltaicos em geral. Seguidamente são apresentados os vários modelos matemáticos que
permitem caracterizar esses sistemas fotovoltaicos bem como os vários métodos utilizados
para estimar os respetivos parâmetros. Essa estimação é feita a partir da informação
disponibilizada pelos fabricantes ou através de dados medidos experimentalmente.
Por conseguinte, é aqui proposto um novo método denominado de Multiswarm Spiral Leader
Particle Swarm Optimization (M-SLPSO), com base no PSO, para resolver o problema da
estimação dos parâmetros dos sistemas PV. Este método utiliza vários swarms com diferentes
mecanismos de procura e cada swarm é guiado por um líder com uma trajetória em espiral
diferente. De acordo com o desempenho dos swarms, estes podem trocar de mecanismos de
procura entre si e os agentes podem migrar entre swarms, possibilitando assim um bom
balanço entre os mecanismos de intensificação e de diversificação. Este método mantém uma
boa diversidade nas trajetórias de exploração enquanto constrói novas soluções ao longo do
processo de procura, mitigando a estagnação da população e a convergência prematura. Para
além disso, consegue explorar o espaço de procura multidimensional em diferentes regiões
simultaneamente e consegue adaptar-se aos vários problemas de otimização.
Finalmente, é feita uma análise e discussão dos resultados obtidos pelo método na resolução
de funções de benchmark e na estimação dos parâmetros fotovoltaicos. Esses resultados são
comparados com os resultados obtidos por vários algoritmos metaheurísticos de última
geração e mostram que o método proposto apresenta um desempenho muito competitivo,
encontrando soluções muito precisas e fiáveis.
The interest in exploring photovoltaic energy has grown a lot in recent years. This is due to the fact that is a type of renewable energy widely available (as it comes from the sun) and has great potential for improvement and expansion. To this end, it is essential to develop methods for predicting and analyzing the behavior of these photovoltaic systems under any operation conditions, i.e., methods that estimate with precision the photovoltaic model parameters under any operation conditions. The main goal of this dissertation is to accurately describe the modeling of photovoltaic systems. Firstly, a presentation is made about the types of photovoltaic technologies currently available and also about the operation of the photovoltaic systems in general. After that, various mathematical models that allow the characterization of these photovoltaic systems as well as the various methods used to estimate their parameters are also presented. This estimation is either based on information provided by manufacturers or experimentally measured data. Therefore, a new method called Multiswarm Spiral Leader Particle Swarm Optimization (MSLPSO) based on the PSO, is proposed to solve the PV parameter estimation problem. This proposed method uses several swarms with different search mechanisms and each swarm is guided by a leader with a different spiral trajectory. Depending on the performance of the swarms, they can exchange search mechanisms with one another, and agents can migrate between swarms, enabling a good balance between the intensification and diversification mechanisms. This method maintains a good diversity in the exploration trajectories while building new solutions throughout the search process, mitigating population stagnation and premature convergence. In addition, it can explore the multidimensional search space in different regions simultaneously and can adapt to various optimization problems. Finally, an analysis and discussion of the obtained results by the algorithm in the resolution of benchmark functions and in the estimation of photovoltaic parameters is made. These results are compared with the results obtained by several state-of-the-art metaheuristic algorithms and they show that the proposed algorithm presents a very competitive performance, finding highly accurate and reliable solutions.
The interest in exploring photovoltaic energy has grown a lot in recent years. This is due to the fact that is a type of renewable energy widely available (as it comes from the sun) and has great potential for improvement and expansion. To this end, it is essential to develop methods for predicting and analyzing the behavior of these photovoltaic systems under any operation conditions, i.e., methods that estimate with precision the photovoltaic model parameters under any operation conditions. The main goal of this dissertation is to accurately describe the modeling of photovoltaic systems. Firstly, a presentation is made about the types of photovoltaic technologies currently available and also about the operation of the photovoltaic systems in general. After that, various mathematical models that allow the characterization of these photovoltaic systems as well as the various methods used to estimate their parameters are also presented. This estimation is either based on information provided by manufacturers or experimentally measured data. Therefore, a new method called Multiswarm Spiral Leader Particle Swarm Optimization (MSLPSO) based on the PSO, is proposed to solve the PV parameter estimation problem. This proposed method uses several swarms with different search mechanisms and each swarm is guided by a leader with a different spiral trajectory. Depending on the performance of the swarms, they can exchange search mechanisms with one another, and agents can migrate between swarms, enabling a good balance between the intensification and diversification mechanisms. This method maintains a good diversity in the exploration trajectories while building new solutions throughout the search process, mitigating population stagnation and premature convergence. In addition, it can explore the multidimensional search space in different regions simultaneously and can adapt to various optimization problems. Finally, an analysis and discussion of the obtained results by the algorithm in the resolution of benchmark functions and in the estimation of photovoltaic parameters is made. These results are compared with the results obtained by several state-of-the-art metaheuristic algorithms and they show that the proposed algorithm presents a very competitive performance, finding highly accurate and reliable solutions.
Description
Keywords
Estimação de Parâmetros Modelo de Dois Díodos Produção Fotovoltaica Um Díodo