| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 3.88 MB | Adobe PDF |
Autores
Resumo(s)
Os modelos de simulação energética em edifícios assumem um papel fundamental na previsão do consumo energético, na avaliação das poupanças de energia obtidas pela aplicação de medidas de melhoria e na certificação energética. Para se ter elevado grau de confiança, é essencial que o modelo represente de forma rigorosa o comportamento atual do edifício em estudo, sendo por isso necessário calibrá-lo de modo a reduzir as diferenças encontradas entre os resultados da simulação e da medição. Estas diferenças são resultado do grande número de parâmetros que definem um edifício e da respetiva incerteza, como por exemplo as características da envolvente e dos sistemas técnicos instalados e sobretudo os perfis de utilização, que têm grande influência no resultado da simulação. A natureza estocástica dos ocupantes e a sua interação com o edifício é ainda tipicamente condensada em perfis homogéneos e pré-definidos, ignorando a diversidade e interdependência dos diversos comportamentos existentes ao longo do dia, levando por isso a consumos simulados bem distintos dos consumos reais. A presente dissertação propõe uma metodologia para se obterem perfis de utilização, de iluminação e de equipamentos, para edifícios de escritórios, com o objetivo de minimizar as diferenças verificadas entre a medição e a simulação. Esta metodologia tem por base a variação horária e diária dos perfis sendo que através da utilização de um algoritmo de otimização (algoritmo genético - NSGA-II) definido por duas funções objetivo, que avaliam as diferenças registadas entre consumos simulados e medidos (MBE e CVRMSE), é possível selecionar os perfis que conduzem à calibração do modelo. Apesar desta metodologia permitir reduzir os erros associados aos perfis de utilização, apenas é verificada a condição de calibração do modelo de simulação (|𝑀𝐵𝐸|≤10% e 𝐶𝑉𝑅𝑀𝑆𝐸≤30%) quando esta é aliada a um conhecimento geral do comportamento dos ocupantes do edifício.
Building energy simulation models play a key role in predicting energy consumption as well as evaluating the extent to which improvement strategies contribute to energy savings. However, in order to be reliable, simulation models need to display the current behaviour of each building as accurately as possible. In other words, they need to be calibrated to reduce the gap between simulation and measurement outcomes. The difference between simulation and measurement outcomes emerge from numerous parameters of buildings. These parameters include the features of buildings envelope, local operating systems and user profiles, with the latter contributing the most to simulation outcomes. The stochastic nature of its users and how they interact with the building tends to be shaped by homogenous and pre-determined profiles. These profiles overlook the diversity and complexity of how individuals behave in a building daily, leading to unrealistic estimated usage of energy. The present thesis proposes a methodology that allows the development of user profiles, lighting and electrical devices, for office buildings, with the aim to reduce the gap between simulation and measurement outcomes. The methodology is based on hourly and daily variations of profiles according to an optimisation algorithm (the genetic algorithm NSGA-II). This algorithm comprises two objective functions which measure the differences between simulated and measured outcomes (MBE and CVRMSE) and which allow the selection of profiles that contribute to the calibration of the model. Even though this methodology contributes to a significant reduction in the errors associated with user profiles, the simulation model can only be calibrated (|𝑀𝐵𝐸|≤10% e 𝐶𝑉𝑅𝑀𝑆𝐸≤30%) when there is a general understanding of how individuals behave within the building.
Building energy simulation models play a key role in predicting energy consumption as well as evaluating the extent to which improvement strategies contribute to energy savings. However, in order to be reliable, simulation models need to display the current behaviour of each building as accurately as possible. In other words, they need to be calibrated to reduce the gap between simulation and measurement outcomes. The difference between simulation and measurement outcomes emerge from numerous parameters of buildings. These parameters include the features of buildings envelope, local operating systems and user profiles, with the latter contributing the most to simulation outcomes. The stochastic nature of its users and how they interact with the building tends to be shaped by homogenous and pre-determined profiles. These profiles overlook the diversity and complexity of how individuals behave in a building daily, leading to unrealistic estimated usage of energy. The present thesis proposes a methodology that allows the development of user profiles, lighting and electrical devices, for office buildings, with the aim to reduce the gap between simulation and measurement outcomes. The methodology is based on hourly and daily variations of profiles according to an optimisation algorithm (the genetic algorithm NSGA-II). This algorithm comprises two objective functions which measure the differences between simulated and measured outcomes (MBE and CVRMSE) and which allow the selection of profiles that contribute to the calibration of the model. Even though this methodology contributes to a significant reduction in the errors associated with user profiles, the simulation model can only be calibrated (|𝑀𝐵𝐸|≤10% e 𝐶𝑉𝑅𝑀𝑆𝐸≤30%) when there is a general understanding of how individuals behave within the building.
Descrição
Tese de mestrado integrado, Engenharia da Energia e do Ambiente, Universidade de Lisboa, Faculdade de Ciências, 2018
Palavras-chave
Calibração Simulação energética Perfis de utilização Edifícios de escritórios Algoritmo Genético Teses de mestrado - 2019
