An adaptive fuzzy energy management strategy for fuel cell hybrid vehicles considering soft classification for driving pattern

An effective energy management strategy (EMS) acts as the brain, ensuring the safe and efficient operation of fuel cell hybrid vehicles (FCHVs). Current driving pattern recognition based EMSs typically employ hard classification to categorize driving patterns into predetermined discrete states, but for data near decision boundaries, using soft classification with probabilistic methods can enhance accuracy. This paper proposes an adaptive EMS strategy that leverages soft classification of driving patterns and optimized fuzzy logic controllers (FLCs) to enhance the performance of FCHVs. Fuzzy c-means clustering and neural networks are used to assign fuzzy membership degrees to driving patterns, enabling the soft recognition of these patterns. Corresponding fuzzy rules are developed by analyzing the offline optimal dynamic programming (DP) energy management results for each of the three identified driving patterns. The membership function parameters of the three FLCs are optimized using the grey wolf optimizer to ensure alignment with DP results. The effectiveness of the proposed EMS is evaluated through a composite test driving cycle that includes three typical driving conditions. The results indicate that the proposed method outperforms the other six algorithms by improving fuel consumption, fuel cell efficiency , and lifetime. It reduces energy consumption by at least 0.27% and decreases fuel cell aging by at least 4.48%. • An online adaptive energy management strategy based on soft classification DPR and an optimized FLC is proposed. • A soft classification method based on fuzzy c-means clustering and neural network is designed for DPR. • The GWO refines FLC parameters to learn offline optimal control data from DP. • The presented algorithms show superiority in terms of fuel consumption, fuel cell efficiency, and fuel cell aging.