Applying simulation-based optimization to improve energy efficiency in two generic office buildings

This study was geared at optimizing the applications of low energy technologies in office buildings. Energy and resource saving measures were extracted from existing high performance buildings practices through a subjective survey among building professionals, which provided guidelines for further optimization work on adapting building parameters and components. The two reference buildings were conceived to meet the requirements of the Canadian Model National Energy Code for Buildings. The path taken for optimization divided the problem into three phases. First, TRNSYS models were developed to predict the energy performance of the two buildings, and the simulation outputs were compared to the results in literature for accuracy confirmation. Then, building characteristics and components were varied in the TRNSYS models to build a database, for training and testing Artificial Neural Network (ANN) models for Response Surface Approximations (RSA). Finally, the ANN model was invoked inside Genetic Algorithm loops, in an attempt to search for the best combination of building parameters that could reduce the energy consumption of the target buildings to the most. The final optimization results demonstrated that up to 39% energy saving could be achieved in both buildings by upgrading the building envelop, enhancing the ventilation regulation, reducing lighting power density, and improving the efficiency of electrical appliance and HVAC systems.