Modeling and optimization of different sparse Augmented Firefly Algorithms for ACMV systems under two case studies

This paper examines the six different schemes of sparse Augmented Firefly Algorithm (AFA) for studying the balancing of energy efficiency and indoor thermal comfort of smart buildings. Based on the well-trained Extreme Learning Machines (ELM) and Neural Networks (NN) models of energy consumption, ambient air temperature and air velocity which have earlier been established and validated through experimental studies, our current optimization problem is formulated to associate indoor thermal comfort with energy efficiency of buildings, so that we can evaluate the key parameters that will influence the balancing of these two demands. The optimizations of the objective functions are carried out in real-time by using novel techniques of sparse AFA. We examined six different schemes of AFA, which are different in random-wandering size and random-wandering distribution. This is so that small and large regions with different wandering can be comprehensively studied. Moreover, the Energy Saving Rates (ESRs) of different operating frequencies are predicted through a third order polynomial regression to minimize the Mean Squared Errors (MSE) of the cost functions. Evaluations of the six different schemes show that the scheme named Large Region Gaussian Wandering (LRGW) generally outperforms the others. Given the best experimental results of AFA optimizations and demonstrated through an experimental room, the maximum potential ESR are about −26.5% for Case 1 of general offices and −9.83% for Case 2 of lecture theatres/conference rooms. These are achieved while maintaining indoor thermal comfort in the pre-defined comfort zone.