Liesje Van Gelder

Comparative study of metamodelling techniques in building energy simulation: guidelines for practitioners

Computer simulation of real system behaviour is increasingly used in research and development. As simulation models become more reliable, they also often become more complex to capture the progressive complexity of the real system. Calculation time can be a limiting factor for using simulation models in optimisation studies, for example, which generally require multiple simulations. Instead of using these timeconsuming simulation models, the use of metamodels can be considered. A metamodel approximates the original simulation model with high condence via a simplied mathematical model. A series of simulations then only takes a fraction of the original simulation time, hence allowing signicant computational savings. In this paper, a strategy that is both reliable and time-ecient is provided in order to guide users in their metamodelling problems. Furthermore, polynomial regression (PR), multivariate adaptive regression splines (MARS), kriging (KR), radial basis function networks (RBF), and neural networks (NN) are compared on a building energy simulation problem. We nd that for the outputs of this example and based on Root Mean Squared Error (RMSE), coecient of determination (R 2 ), and Maximal Absolute Error (MAE), KR

Designing uncertain optimization schemes for the economic assessment of stock energy-efficiency measures

There are a myriad of options when it comes to designing an uncertain optimization scheme. This work explores the impacts of several of these options using a case-study retrofitting of Swedish attics developed during the Annex 55 project of the International Energy Agency Energy Buildings and Community programme. We find that not considering maintenance costs, increments in energy costs with time, and discount rates in the economic criterion can impact the derived optimal design option. However, for our case study no difference was found between criteria that examined absolute returns, fractional returns, and annulized returns, despite emphasizing different aspects of the investment. If multiple moments of the economic criterion distribution are combined with weights to create a single objective function then several weights should be explored. Alternatively a single moment could be sought, such as the probability that the economic criterion is negative as a sufficient single objective, avoiding having to choose a weight. A hybrid genetic algorithm may offer a more efficient method of determining the optimal design option but a multi-layered sampling scheme offers more flexibility regarding subsequent analysis.