Factorial design for energy-system models

Mathematical models are extensively used in energy analysis and have increased in scope as better and faster computers have become available. With complicated systems, it is difficult to predict accurate results if doubtful input data are changed. Traditionally, sensitivity analysis with a change of one or more of the parameters is used. If the influence of a change is very small, the first result is believed to be accurate. Problems may arise when sensitivity analysis is applied to a vast amount of data. The aim of this paper is to examine whether the calculation effort can be decreased by using factorial design. Our model, called Opera (Optimal Energy Retrofit Advisory), is used to find the optimal retrofit strategy for a multi-family building. The optimal solution is characterised by the lowest possible life-cycle cost. Three parameters have been studied here: length of the optimisation period, real interest rate and existing U-value for an attic floor. The first two parameters are found to influence the life-cycle cost significantly, while the last is of minor importance for this cost. We also show that factorial analysis must be used with great care because the method does not reflect the complete situation.