Title: PID self-tuning method based on deep belief network and improved firefly algorithm

Authors: Lingzhi Yi; Xiu Xu; Mao Tan; Zongguang Zhang; Weihong Xiao; Lv Fan

Addresses: College of Automation and Electronic Information, Xiangtan University, Xiangtan, Hunan, China ' Hunan Province Multi-Energy Cooperative Control Technology Engineering Research Center, School of Information Engineering, Xiangtan, Hunan, China ' College of Automation and Electronic Information, Xiangtan University, Xiangtan, Hunan, China ' Key Laboratory of Intelligent Computing and Information Processing, Ministry of Education, Xiangtan, Hunan, China ' China Mobile Zhuzhou Electric Limited by Share Ltd., Zhuzhou, Hunan, China ' Willfar Information Technologies Co. Ltd., Changsha, Hunan, China

Abstract: In order to overcome the difficulty of tuning the proportion integration differentiation (PID) parameters, a PID parameter self-tuning method based on the firefly algorithm improved by Newton's law of universal gravitation (LOGFA) and deep belief network (DBN) is proposed. Compared with the FA, LOGFA cannot only maintain the evolutionary advantage of the original algorithm but also can effectively improve the accuracy and convergence ability of the algorithm. The advantage of DBN is to train each layer of neural network separately, which greatly improves the training efficiency and accuracy. The closed-loop PID speed control system of a three-phase asynchronous motor is used as the simulation object for PID parameter self-tuning. The proposed LOGFA-DBN is compared with other three algorithms. Simulation results show that the algorithm combining LOGFA and DBN can realise the off-line parameter tuning which is not subject to the controlled object, and speed up the parameter tuning.

Keywords: proportion integration differentiation; PID; LOGFA; deep belief network; DBN; parameter self-tuning.

DOI: 10.1504/IJAAC.2021.114931

International Journal of Automation and Control, 2021 Vol.15 No.3, pp.363 - 377

Received: 06 Aug 2019
Accepted: 10 Mar 2020

Published online: 12 May 2021 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article