Development of Exponential-Fitted Numerical Methods for Building Energy Simulation

Simulation of energy flows in buildings is computationally intensive. Consequently, improvements in computing power and algorithm efficiency can always be utilized. Commonly used implicit solvers require extensive matrix processing while standard explicit methods have limited stability and progress in very small time increments. In this work, a number of stable numerical methods are examined which are explicit in nature and therefore do not require the use of matrices. One such algorithm and two proposed developments of it are assessed using a building related test problem prepared for this purpose, and their performances are compared with that of an efficient implicit method. The proposed algorithms are found to be the best in their class with computational efficiencies approaching that of the implicit method. In addition, this class of method makes more efficient use of current computer resources and is particularly well suited to future (parallel) architectures.