Power consumption minimization by distributive particle swarm optimization for luminance control and its parallel implementations

• Published in: Expert systems with applications Vol. 96; pp. 479 - 491
• Main Authors: Liao, Chih-Lun, Lee, Shie-Jue, Chiou, Yu-Shu, Lee, Ching-Ran, Lee, Chie-Hong
• Format: Journal Article
• Language: English
• Published: New York Elsevier Ltd 15.04.2018; Elsevier BV
• Subjects: Computer simulation; CUDA; Electricity; Energy conservation; Energy consumption; GPU; Hadoop; Illuminance; Industrial plants; Lighting; Luminance; MapReduce; Parallel algorithm; Parallel processing; Particle swarm optimization; Power consumption; Real time; Working conditions; Workshops; CUDA; Parallel algorithm; Energy conservation; Hadoop; Particle swarm optimization; GPU; MapReduce
• ISSN: 0957-4174; 1873-6793
• DOI: 10.1016/j.eswa.2017.11.002

•Luminance control is formalized as a constrained search problem.•Both power consumption minimization and sufficient illuminance are considered.•A distributive PSO-based algorithm is developed to do an effective search.•Parallel implementations in GPU and Hadoop MapReduce are developed.•The developed systems are demonstrated to be effective in real-time luminance control. We present an intelligent system, based on the particle swarm optimization (PSO) technique, to solve a power consumption minimization problem which is commonly encountered at the industrial factories or workshops. The power minimization problem is concerned with adjusting the settings of a number of lighting devices in real time in a working environment, subject to the requirements of minimizing the power consumption of the lighting devices as well as producing sufficient illuminance over all the specified working spots in the working area. Usually, the search space involved is too huge and solving the problem with traditional methods, e.g., brute force or least squares, is out of the question. In this paper we describe a distributive-PSO (DPSO) based algorithm to solve the problem. We show that by dividing the whole population of particles into a number of groups, PSO can be done distributively on each group and the best settings for the lighting devices, which meet the requirements, can be efficiently obtained. DPSO is very suitable to be parallelized. Parallel implementations in GPU and Hadoop MapReduce are developed. Simulation results show that our developed system is effective for a variety of working environments. We believe our work facilitates developing an efficient tool for energy conservation as well as other optimization applications.