Assessment of Applications of Optimisation to Building Design and Energy Modelling

Buildings account for around 35% of the world’s carbon emissions and strategies to reduce carbon emissions have made much use of building energy modelling. Optimisation techniques promise new ways of achieving the most cost effective and efficient solutions more quickly and with less input from engineers and building physicists. However, there is limited research into the practical applications of these techniques to building design practice. This thesis presents the results of case-based research into the practical application of design stage optimisation and calibration methods to energy efficient building fabric and services design using building energy modelling. The application during early stage design of a Non-dominating Sorting Genetic Algorithm 2 (NSGA2) to a building energy model EnergyPlusTM. The exercise was used to determine if the application of NSGA2 yielded a significant improvement in the selection of building services technology and building fabric elements. The use of NSGA2 enabled significant (£400,000) capital cost savings without degrading the comfort or energy performance. The potential capital cost savings significantly outweighed the cost of the engineering time required to carry out the additional analysis. Three optimisation techniques were applied to three case study buildings to select appropriate model parameters to minimise the difference between modelled and measured parameters and hence calibrate the model. An heuristic approach was applied to the Institute for Life Sciences Building 1 (ILS1) at Swansea University. Latin Hypercube Monte Carlo (LHMC) was applied to the Arup building at 8 Fitzroy St London and compared directly with the results from an approach using Self Adaptive Differential Evolution (SADE). Poor Building Management System data quality was found to significantly limit the potential to calibrate models. Where robust data was available it was however found to be possible to calibrate EnergyPlus simulations of complex real world buildings using LHMC and SADE methods at levels close to that required by professional bodies.