Online optimal control of multiple-chiller systems in large buildings

Abstract

This thesis investigates the online optimal control of multiple-chiller systems in large buildings with enhanced robustness and cost efficiency. The control mainly includes chiller sequencing control, optimal start control and electrical demand limiting control. New strategies in the three subjects are proposed and validated on a dynamic simulation platform as well as using site data. In the first subject, i.e. chiller sequencing control, three methods are developed to enhance its robustness and reliability. Firstly, the data fusion method is proposed to obtain a more accurate and reliable cooling load measurements by combining the complementary advantages of two different load measurements. Secondly, a simplified model is developed to online compute the varying maximum cooling capacity of individual chiller. Thirdly, an online sensor fault detection and diagnosis (FDD) strategy is developed to ensure that the sensors used in the direct measurement III work healthily. In the optimal start control, a model-based strategy is proposed for minimizing the energy consumption of the central chilling plant in the morning start period. The model-based strategy is realized in two steps. The first step is to identify a feasible set for the operating chiller number. The second step is to estimate the pre-cooling lead time using the simplified building model for each number inside the feasible range identified, and calculate the corresponding energy consumption. In the subject of electrical peak demand limiting control, a strategy of minimizing the monthly electricity bill is proposed via utilizing the building thermal mass. Previous studies can not to achieve maximized monthly cost saving because the demand cost reduction may be largely/completely traded off by the energy cost rise. Therefore, a strategy taking full consideration of the relationship is developed. The strategy consists of two phases. The first one is to predict a suitable monthly peak demand threshold. In the second phase, the extended pre-cooling lead time will be determined based on the difference between the demand threshold and the predicted daily peak demand. The developed strategies are validated through case studies, which show the satisfactory performances.