A Model-free Approach to Optimal Chiller Loading Problem Using Global Simultaneous Perturbation Stochastic Approximation

Abstract

This paper presents an investigation of a modelfree approach using global simultaneous perturbation stochastic approximation (GSPSA) to optimal chiller loading problem. The GSPSA based method is employed to optimize the partial load ratio (PLR) of each chiller such that the total power consumption of multi-chiller systems is minimized. The main advantage of GSPSA is that it can produce fast design parameter without information of plant model by measuring the input-output data of the system. Our model-free design is validated using an experimental data from a well-known multichiller system of semiconductor factory in Hsinchu Scientific Garden, Taiwan. In addition, the performance of the GSPSA based method is compared to the other stochastic optimization based approaches. Simulation results illustrate that the GSPSA based method has a potential in minimizing power consumption of multi-chiller systems with less number of evaluated cost functions.