A parallel solution with GPU technology to predict energy consumption in spatially distributed buildings using evolutionary optimization and artificial neural networks. (1st March 2020)

• Record Type: Journal Article
• Authors: Iruela, J.R.S.; Ruiz, L.G.B.; Pegalajar, M.C.; Capel, M.I.
• Is Part Of: Energy conversion and management. Volume 207(2020)
• Publication Date: 2020-03-01
• Subjects: Energy consumption forecasting -- Artificial neural networks -- GPU -- Evolutionary algorithm; Direct energy conversion -- Periodicals; Energy storage -- Periodicals; Energy transfer -- Periodicals; Énergie -- Conversion directe -- Périodiques; Direct energy conversion; Periodicals; 621.3105
• DOI: 10.1016/j.enconman.2020.112535
• Languages: English
• ISSNs: 0196-8904

Abstract

Highlights: A GPU-based approach to forecast energy consumption in buildings. An effective implementation of NSGA-II and ANN based on parallel optimization. GPU design far outperforms the classical implementation in terms of time. Abstract: Today all governments talk about climate change and its consequences. One of the ways to tackle this problem is by studying the energy consumption of the buildings around us. The study of energy consumption may give us relevant information to make better decisions, and thus reduce costs and pollution. However, ANNtraining models, in order to achieve those goals, has a high computational cost in terms of time. To solve that problem, this paper presents a GPU-based parallel implementation of NGSA-II to train ANNs whose evaluation has also been implemented in a parallel GPU scheme. Our methodology is designed to predict the energy consumption of a series of public buildings, and thus, to model consumption, save energy and improve the energy efficiency of these buildings without compromising their performance obtaining the prediction in a very short period of time. We compared the sequential implementation of the evolutionary algorithm NSGA-II with our new version developed in parallel and the parallel implementation gets better results in much faster execution time.