Optimizing energy management strategy for fuel cell hybrid electric vehicles: A hybrid FBPINN-MGO Approach

Traditional energy management strategies (EMS) for fuel cell hybrid electric vehicles (FCHEVs) often overlook driving cycle uncertainties from fluctuating traffic, impacting fuel economy, battery life, and system efficiency. This paper proposes a hybrid approach for optimizing FCHEVs by combining Finite Basis Physics-Informed Neural Networks (FBPINN) with the Mountain Gazelle Optimizer (MGO), also known as the FBPINN-MGO method. The goal is to maximize fuel economy, extend EV range, and prolong battery life by optimally distributing power between the fuel cell and battery for peak performance. The FBPINN model is used to predict the fuel economy of the FCHEV motor controller, while the MGO algorithm is utilized to optimize the control parameters for fuel economy. The MATLAB platform is employed to develop the proposed FBPINN-MGO method, and its performance is compared to that of other existing methods, such as the Deep Neural Network (DNN), Elman Neural Network (ENN), and Bayesian Regularization Neural Network (BRNN). The proposed FBPINN-MGO method achieves 95 % efficiency, 0.632 % error, and 0.954 % sensitivity, and lowers the computational cost to 0.0530 $/km, outperforming existing methods. The proposed method enhances efficiency, optimizes battery life, minimizes error propagation, and adapts to dynamic conditions for a robust FCHEV EMS.