Abstract
A robust and efficient nonlinear algebraic and differential equations solver is increasingly essential as the scale of dynamic building system simulation grows larger and the models employed become more complex. There are many factors affecting the performance of these solvers. In this paper the impact of variable scaling on the convergence of a family of Newton-method based solvers is summarized. A comparison of the performance of a few different scaling methods is presented using HVACSIM+, a software simulation tool that employs Powell’s Hybrid method to solve the system of nonlinear algebraic equations. A small commercial building is modelled in HVACSIM+ and serves as an illustrative example for this study. It is demonstrated that, for these methods, variable scaling can strongly influence computational performance but, for our example problem, any improvement or deterioration in numerical accuracy is limited. The conclusions are not limited to the software platform HVACSIM+.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call