Presentation Details
| Physics-Informed TCN Transformer with Localized P&O Refinement for Rapid Global MPPT under Partial Shading Natnael Dejene, Sandip Das. Kennesaw State University, Marietta, GA, USA |
Abstract
Partial shading in photovoltaic (PV) arrays creates multiple local maxima in the power-voltage characteristics, challenging traditional maximum power point tracking (MPPT) algorithms. This paper presents a hybrid framework combining a physics-informed Temporal Convolutional Network (TCN) and Transformer architecture with a localized dynamic-step Perturb and Observe (P&O) refinement for rapid global MPPT (GMPPT). The network extracts sequential and global features to predict coarse GMPP estimates while a physics-informed loss ensures adherence to the PV incremental conductance law. The hybrid system achieved a mean terminal tracking efficiency of 99.937% under severe partial shading, with a mean power tracking error of 0.062% and convergence within 208.4 ms, demonstrating both high accuracy and real-time applicability. This work provides a scalable, edge-deployable solution for next-generation highly efficient PV systems.
No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the author.
No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the author.
Content Locked. Log into a registered attendee account to access this presentation.