EnerMorph-net: energy state morphing hyper network for self-optimizing IoT-driven electric vehicle systems

Abstract Aim An adaptive deep learning architecture, termed EnerMorph-Net, is proposed to overcome the limitations of conventional models in Internet of Things (IoT)-enabled electric vehicle transportation systems. The framework captures dynamic energy state variations influenced by traffic density, weather conditions, and power grid fluctuations thereby enabling accurate energy state prediction and intelligent decision-making for efficient electric vehicle energy management. Findings The proposed EnerMorph-Net shows better performance than the traditional CNN and LSTM models. It attains an MAE of 0.066, RMSE of 0.091, MAPE of 6.84%, and R² score of 0.936 which is equal to 8.7% better accuracy in prediction. In addition to it, the model ensures a decrease in energy variance by 11.3%, charging cost efficiency by 9.5%, and grid interaction stability by 14.2%. Furthermore, the model also shows an increase of 6.8% in robustness towards sensor noise and missing IoT data.