Evolutionary Synthesis of HVAC System Configurations: Experimental Results

The aim of this research was to investigate the synthesis of novel heating, ventilating, and air conditioning (HVAC) system configurations using model-based optimization methods (Wright et al. 2008). This paper describes the experimental results for the optimization of a two-zone HVAC system of a building located in a continental climate. The goal of the optimization was to find a feasible system design that operated with the minimum system capacity at each load condition. The optimization method used in this research is based on a Genetic Algorithm search method. The robustness of the optimization was examined through the consistency of the design solutions found from multiple runs of the algorithm (each run being subject to different initial conditions). The results indicate that given two runs of the algorithm, there was a high probability of finding a system design that has a performance comparable to existing system configurations. Given eight runs of the algorithm, it is probable that the best system found would have a performance that exceeded that of existing system configurations. However, approximately one third of all optimization runs would converge onto an infeasible system configuration, the elimination of this characteristic being the subject of future research. The optimality of the synthesized systems was judged in comparison to the performance of three benchmark systems and by comparing the system capacity to the minimum possible at a given load condition. The best of the synthesized systems had a performance that exceeded that of the conventional benchmark systems and that was comparable to that of a conceptually optimum system configuration. The system capacity was also close to the minimum possible capacity and as such was judged to be a near-optimum system configuration for the example building. It can be concluded that the optimization approach is able to synthesize near-optimum system configurations that have a performance equal to or better than existing system configurations. The algorithm, however, requires multiple runs in order to find reliable solutions, a fact that should be addressed in future research. The current algorithm, however, represents a significant step toward the design of software systems that are able to synthesize new and optimum HVAC system configurations.