Optimizing system control with load prediction by neural networks for an ice-storage system

Abstract

This paper describes the performance of a partial ice storage system that has a controller that predicts the load by neural networks. To evaluate the performance, a comparison was carried out between the two control strategies--chiller priority control and predictive control--using simulation. Chiller priority is the most common control strategy for existing thermal storage systems. The predictive control proposed in this study uses an hourly thermal load prediction by neural networks. The predictive control is described in detail. The study indicates that the accuracy of the load prediction is a key for optimizing the system control. The predictive control can significantly reduce the operating cost without energy shortage.