Intelligent system based supervision for energy management of water chiller plant

For almost a decade, intelligent systems have shown great potential for solving non-linear control problems. This paper exhibits a possibility of using a neuro-fuzzy system to advise a chiller plant. This plant, located at Building No.11 of Rangsit University, consists of three water chillers. Nonlinear dynamic characteristics, resulting from outdoor conditions, and equipment within the plant such as pumps and modulation valves, are discussed herein. The objective of this research is to minimize energy cost while maintaining high performance of the plant. In this paper, identification and prediction of a non-linear discrete time system using a neuro-fuzzy system is investigated. To develop an algorithm for achieving such an objective, the first step is to learn the relationship of outdoor temperature, humidity, and cooling load of the building, or the plant dynamics, from historical data. After being trained completely, the neuro-fuzzy system is used to predict the cooling load for Building No.11 at Rangsit University and operating optimal points of the water chiller plant. We use the visual studio to develop an advised chiller plant management program. This program learns the information from both historical and presents data of the chiller plant. After that, we compared energy costs before (May-July 2014) and after (August-October 2014) using advisory management of the chiller plant program. The system can reduce electrical cost within 5%. Furthermore, it can be trained by plant operators continuously.