Optimizing high voltage gain interleaved boost converters for PV and wind systems using hybrid deep learning with bitterling fish and secretary bird algorithms

Abstract

The integration of renewable energy sources such as photovoltaic (PV) and wind systems demands high-efficiency, high-voltage gain power conversion architectures. However, interleaved boost converters, while suitable for such applications, face challenges in balancing complexity, scalability, cost, and dynamic environmental variability. This study introduces a series of novel intelligent control frameworks to overcome these limitations and improve overall system performance. Firstly, a Neuro-LSTM BitterSec Optimization Network (NL-BSONet) is proposed to enhance the efficiency of high-voltage gain interleaved boost converters while minimizing system complexity. This hybrid approach leverages neural networks and LSTM-based learning for real-time optimization, offering improved scalability and lower switching losses. To address power quality issues caused by fluctuating irradiance and wind speeds, the study introduces the Adaptive Neuro-Deep Reinforcement Learning Bitterling Optimizer (AN-DRLBO). This model integrates Deep Reinforcement Learning (DRL) for adaptive energy conversion, Adaptive Neural Networks (ANN) for real-time system stabilization, and Bitterling Fish Optimization (BFO) for robust performance under transient conditions. Furthermore, due to the difficulty in achieving optimal control parameters under variable environmental conditions, an Adaptive LSTM-Encoded Secretary Optimization Network (AL-SONet) is developed. This framework employs Long Short-Term Memory (LSTM) networks for predictive control, Autoencoder-based Optimization (AEO) for feature extraction and simplification, and Secretary Bird Optimization (SBO) for dynamic parameter tuning. The proposed architectures demonstrate superior performance, achieving 97 % energy conversion efficiency, a voltage gain of 32.5 dB, and minimal output ripple, thereby ensuring stable and efficient integration of renewable energy sources. This research contributes a comprehensive and adaptive control solution for next-generation renewable energy systems.