Title: Neural network observer-based fractional-order linear quadratic control of modular multilevel converter

Authors: Vivek Patel

Addresses: Madan Mohan Malaviya University of Technology, Deoria Road, Gorakhpur, Uttar Pradesh, 273010, India

Abstract: The modular multilevel converter (MMC) has seen extensive application in the high-voltage sector. The MMC is composed of multiple similar submodules. To achieve optimal performance, the MMC must balance the voltages between its upper and lower capacitor arms while suppressing alternating current components in the circulating current. This paper proposes a novel approach for controlling MMCs using a neural network observer-based fractional-order linear quadratic integral control technique. By leveraging neural network observers, the system is able to estimate internal states, while the fractional-order control strategy ensures robust performance and enhanced stability in the face of varying operating conditions. The effectiveness of the proposed method is demonstrated through simulations (MATLAB 2023a), highlighting its ability to regulate key converter parameters such as load current, circulating current, and capacitor voltages. The simulation results show the superior performance in comparison to other controller approach, particularly in scenarios with fluctuating supply voltage.

Keywords: modular multilevel converter; MMC; neural network observer; fractional-order; time varying linear quadratic control.

DOI: 10.1504/IJPELEC.2025.149580

International Journal of Power Electronics, 2025 Vol.21 No.6, pp.543 - 559

Received: 14 Nov 2024
Accepted: 15 Mar 2025
Published online: 07 Nov 2025 *

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