A rational choice of solutions for low energy building requires a performance evaluation for a large number of parameter combinations. Currently, building thermal dynamic simulation is used to check design solutions but it gives little information on what and how much may be improved. Iterative parameter optimization is a solution but the methods used may stick in local minima and do not allow the designer to evaluate different solutions. An alternative evaluation method is to obtain simple polynomial functions which estimate the annual energy demand as a function of building envelope parameters. The coefficients of these functions may be obtained by regression from dynamic simulation results. The number of dynamic simulations needed is reduced if the numerical simulations are optimally designed. The polynomial functions may be used to predict alternative performances from which the solution may be chosen. The designer is guided in his choice by the coefficients of the polynomial function which give the effect of each parameter on energy demand. (C) 2009 Elsevier B.V. All rights reserved.