The Equivalent Consumption Minimization Strategy (ECMS) is an effective approach for managing energy flow in hybrid electric vehicles (HEVs), balancing the use of electric energy and fuel consumption. The strategy’s performance depends heavily on the Equivalent Factor (EF), which governs this trade-off. However, the optimal EF varies under different driving conditions and is influenced by the inherent randomness in factors such as traffic, road gradients, and driving behavior, making it challenging to determine through traditional methods. This paper introduces Bayesian Optimization (BO) as a solution to address the stochastic nature of the EF parameter tuning process. By using a probabilistic model, BO efficiently navigates the complex, uncertain performance landscape to find the optimal EF parameters that minimize fuel consumption and emissions across variable conditions. Simulation results under WLTP cycles show that the proposed method reduces fuel consumption by 0.9% and improves SoC control accuracy by 0.3% compared to conventional ECMS. Compared to rule-based EMS, it achieves an 8.2% fuel economy improvement and a 5.3% reduction in SoC error. Simulation results for the WLTP demonstrate that the EF parameters optimized with BO lead to improved ECMS performance, reducing fuel consumption and enhancing energy efficiency. The method also achieves the lowest SoC variance under uncertainty, demonstrating its robustness and practical applicability. This study highlights the advantages of Bayesian Optimization in tuning ECMS for hybrid vehicles, offering a robust approach to handle the uncertainties and variability in driving cycles, ultimately contributing to more efficient and sustainable vehicle energy management.