Energy management systems (EMS) regulate fuel consumption in battery and fuel cell (FC) vehicles. However, fuel cells face challenges like limited hydrogen infrastructure, and high costs. This a hybrid method that combines the Lyrebird Optimisation Algorithm (LOA) and Self-Attention Generative Adversarial Networks (SAGAN), referred to as The LOA-SAGAN technique. This proposed method's major purpose is to lessen fuel costs and improve system performance. Additionally, battery degradation is prioritized. The fuel usage is decreased. The proposed LOA is used to optimize the dc–dc converter's parameters and the SAGAN is utilized to predict the optimized parameters. The proposed approach is put into practice on the MATLAB platform. The proposed strategy yields superior outcomes in all existing systems like Fuzzy logic Control (FLC), Genetic Algorithm (GA) and the Multi-Island Genetic Algorithm (MIGA). The proposed technique indicates a cost of $1320, fuel consumption of 17.9 g, an accuracy of 96%, and a sensitivity of 97%. The LOA-SAGAN technique effectively reduces costs and fuel consumption while improving energy management performance.