Parameter Estimation and Optimal Supervisory Control of Chilled Water Plants

Abstract

The objective of the work is to develop methods for minimizing the energy costs of chilled water systems through optimal control. The general approach includes the following tasks: 1) proposing a system of parametric models that represent the real chilled water system, 2) determining model parameters from measured data, and 3) subjecting the system of parametric models to an optimization algorithm. A comprehensive approach for determining the optimal control for any general chilled water system is developed. The general system presents a difficult problem for parameter estimation and optimization because of discontinuous variables and nonlinear relationships between input and output variables. Various methods for parametric estimation and control optimization are pre- sented and demonstrated on simulated and actual plant models. The actual plant model consists of interconnected component models, including an electric motor driven chiller, a steam turbine driven chiller and associated steam condenser, and a multi- cell cooling tower. Different parameter estimation methods using measured plant data are applied and compared. Optimal supervisory control is determined through application of the simulated annealing method to the model. The dependence of optimal control settings upon independent variables (e.g. chilled water return temperature and ambient wet bulb temperature) is investi- gated. Cost savings of optimal over conventional control strategies are calculated and compared.