Fractional factorial design for energy systems

Nowadays, when powerful computers are on almost every one's desk, it has become more and more common to use complex energy-system models in order to predict the use of electricity and heat in buildings. At the same time, it has been harder to grasp the overall solution because of all the details implemented in such a model. A method that could help the operator to find the important parts in the model would therefore be of great interest. Traditionally this is addressed by using so-called sensitivity analyses. The most common method is then to change one input parameter a certain amount and study how much the output is influenced by this change. If the output varies significantly the parameter is supposed to be of more interest than if there is a only small change. If there is a complex model, several hundred parameters may have to be changed this way; this is very tedious. By the use of modern statistics, these calculations can be made in a more planned way and the necessary work minimized. One such method is fractional factorial design, which is used for examining a widely used Swedish energy-balance program with about 70 input data values. We have examined nine of these parameters in order to rank their importance for the output energy balance. The interactions between these nine parameters have also been studied using the same method.