Towards sustainable grids: deep learning-driven prediction and optimisation for EV charging infrastructure and bifacial PV module-based DG in mesh distribution networks to minimise losses and enhance efficiency

The future of transportation is undergoing a transformative shift towards electric vehicles (EVs), driven by their environmental benefits and potential to significantly reduce greenhouse gas emissions. However, the large-scale integration of EVs presents critical challenges, including increased network power losses, voltage instability and heightened generation demands. To address these challenges, this study proposes a novel integrated framework that optimally places and sizes electric vehicle charging stations (EVCS) alongside bifacial photovoltaic (PV) solar-based distributed generation (DG) systems to enhance grid stability and efficiency. Unlike existing approaches, this work leverages an advanced hybrid methodology combining deep learning-based forecasting and metaheuristic optimisation. Specifically, Long Short-Term Memory (LSTM) networks are employed for highly accurate EV charging load and solar irradiance predictions, outperforming conventional methods. Additionally, a bat optimisation algorithm is innovatively applied to optimise EVCS and DG placement, aiming to minimise active/reactive power losses, reduce voltage deviations and maximise voltage stability. The proposed approach is rigorously validated on an IEEE-33 and IEEE-69 bus mesh distribution system under various configurations, DG-only, EV-only and combined DG-EV scenarios, demonstrating substantial improvements in power loss reduction and voltage profile enhancement.