期刊名称：Advances in Electrical and Computer Engineering
印刷版ISSN：1582-7445
电子版ISSN：1844-7600
出版年度：2019
卷号：19
期号：2
页码：13-22
DOI：10.4316/AECE.2019.02002
出版社：Universitatea "Stefan cel Mare" Suceava
摘要：This article presents a new methodology to implement a computational parallel scheme based on Differential Evolution (DE) algorithm through the use of Graphical Processing Units (GPU). A system application in which it is possible to perform an online monitoring of Power Quality Disturbances (PQD) in electric grids is presented as a case study, where a fitting of the parameters of a mathematical model is performed through this technique. Hyper-parameter optimization of the parallel Differential Evolution algorithm is performed for the assigned fitting function. As a result of this parallel implementation, a speed-up of 37 times compared with the serial implementation is achieved by using a single low budget GPU. The work presented shows a significant speed and accuracy improvement compared with Micro-Genetic Algorithm for Power Quality Analysis (MGA-PQA) technique.
