SA-12-010 -- Simulating Central Chiller Plant Loop Delta-T with Quadratic Regression Models

Loop delta-T is a critical parameter for the simulation andoptimization of a central chilled water system, especially forthose with chilled water storage tanks and those with highpump power installation. This paper proposes to model theloop delta-T using a quadratic regression model. The modelingprocedure is divided into three steps: variables selection andfiltering, best regression screening, and model adequacycheck. Eight variables are selected as candidates of significantvariables. The variables with less significance are screenedout using P-value criteria or statistic criteria R_adj and RMSE.The remaining variables are combined with their correspondingcross-product terms to form quadratic regression models.The best quadratic model is identified using Mallows Ckstatistic. Finally, the adequacy of the model is checked withresidual analysis and influential point examination. Thisprocedure was illustrated using data from two real chilledwater systems and reasonably accurate models were obtained.