Title: Adaptive neural networks for AC voltage sensorless control of three-phase PWM rectifiers

Authors: Adel Rahoui; Hamid Seddiki; Ali Bechouche; Djaffar Ould Abdeslam

Addresses: L2CSP Laboratory, Mouloud Mammeri University, BP 17 RP, 15000 Tizi-Ouzou, Algeria ' L2CSP Laboratory, Mouloud Mammeri University, BP 17 RP, 15000 Tizi-Ouzou, Algeria ' L2CSP Laboratory, Mouloud Mammeri University, BP 17 RP, 15000 Tizi-Ouzou, Algeria ' MIPS Laboratory, University of Haute-Alsace, 61 rue Albert Camus, 68093 Mulhouse, France

Abstract: In this paper, a new adaptive grid voltages estimator for AC voltage sensorless control of three-phase pulse-width modulation (PWM) rectifier is proposed. This method is based on a simple adaptive neural network (ANN) to estimate online the grid voltages. Its main advantages are the simplicity and low computational cost requirements. The proposed ANN estimator is inserted in voltage-oriented control (VOC) to perform an AC voltage sensorless control scheme. As the start-up is a common problem in case of sensorless control, a new start-up process is also proposed for estimating initial values of the grid voltages. Experimental tests are carried out to verify the feasibility and robustness of the proposed ANN estimator. Obtained results show good performances of the proposed AC voltage sensorless control scheme in normal and severe grid voltage conditions.

Keywords: adaptive neural network; ANN; adaptive neural filter; ANF; pulse-width modulation; PWM; rectifier; diode rectifier; grid voltages estimation; voltage-oriented control; VOC; sensorless control; start-up process; experimental verification; stability.

DOI: 10.1504/IJMIC.2019.097985

International Journal of Modelling, Identification and Control, 2019 Vol.31 No.2, pp.139 - 151

Received: 10 Aug 2016
Accepted: 30 Jan 2017
Published online: 26 Feb 2019 *

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