Enhancing Power Quality in Smart Grid‐Connected Renewable Energy Systems Using A Hybrid Deep Learning‐Based DSTATCOM: Self‐Improved Jellyfish Optimizer (Si‐Jo)

ABSTRACT This manuscript introduces a robust stochastic optimization technique for real‐time energy management in household solar (PV) storage systems, which addresses the constraints provided by the stochastic nature of solar power and load demand. Traditional offline solutions for day‐ahead Battery Energy Storage (BES) scheduling frequently result in real‐time energy losses, while online alternatives do not adequately reduce electricity prices. The proposed system combines an offline optimization model with a real‐time rule‐based controller and a deep learning‐based Conditional Self‐Attention Generative Adversarial Network (CSAGAN) for load and solar generation predictions to address these constraints. The War Strategy Optimization (WSO) Algorithm, which operates inside a receding horizon framework, optimizes BES dispatch set‐points to lower daily electricity costs. The research also introduces a DSTATCOM‐based hybrid technique for improving power quality (PQ) in grid‐connected solar and wind systems, which makes use of a hybrid deep learning model consisting of Deep Belief Networks (DBNs) and Multilayer Perceptrons (MLPs). The Self‐Increased Jellyfish Optimizer (SI‐JO) generates control signals and optimizes the six‐leg voltage source converter (VSC) and the buck‐boost converter. MATLAB simulations show considerable increases in energy efficiency, harmonic distortion reduction, and system stability under various irradiance and grid fault scenarios, demonstrating the efficacy of the proposed techniques.