Treffer: A Novel Meta‐Learning‐Based Reinforcement Controller for Voltage Regulation of an Interleaved Boost Converter in Solar Photovoltaic Systems.

The increasing integration of solar photovoltaic (PV) systems into modern power grids highlights the need for advanced control strategies that ensure reliable voltage regulation under variable operating conditions. PV output is highly sensitive to irradiance and temperature fluctuations, which can degrade power quality and compromise grid stability. Conventional reinforcement learning‐based controllers, such as deep deterministic policy gradient (DDPG), have shown promise, but their reliance on fixed hyperparameters limits adaptability, leading to performance deterioration during rapid solar variations. This paper proposes a novel adaptive meta‐learning‐based DDPG (AM‐DDPG) controller implemented with a three‐leg interleaved DC–DC boost converter for PV voltage regulation. The proposed controller employs a meta‐learning mechanism to dynamically adjust key hyperparameters, including learning rate, stability factor, and noise scale, thereby improving responsiveness and adaptability. MATLAB simulations compare AM‐DDPG with standard DDPG under slow, fast, and highly variable irradiance and temperature profiles. Results demonstrate that AM‐DDPG achieves voltage stabilization within 10 ms, maintains a 566 V output with less than 0.1% deviation, and significantly suppresses voltage ripples. By enhancing dynamic performance and robustness, the proposed approach supports higher PV penetration and improves conversion efficiency. It also strengthens grid integration of renewable energy, contributing to sustainable and resilient low‐carbon power systems. [ABSTRACT FROM AUTHOR]

Copyright of IET Renewable Power Generation (Wiley-Blackwell) is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)