The supervisory controller discussed is a personal-computer-based software package that operates in conjunction with a building's energy management and control system to optimise cooling equipment operation. The system provides decision support to determine when to operate cooling equipment to minimise costs under real-time pricing or conventional time-of-use electric rates. An artificial neural network (ANN) is used to model the dynamic behaviour of the building and energy equipment while an evolutionary-based search routine, a genetic algorithm (GA), is used for optimisation. A prototype system is installed and operating at a high school in southern California to control a thermal energy storage system, a conventional screw-type chiller, and a gas-fired, engine-driven chiller. Describes the neural network and genetic algorithm and discusses how the two have been integrated into a working system. Emphasis is placed on lessons learned during the developmental process and on insights gained in the practical application of both GA and ANN technologies.

KEYWORDS: year 1997, Computer controls, optimisation, energy management, controls, expert systems, algorithms, prototypes, accuracy, performance, schools, USA, case studies, buildings