Scenario driving cycle development by fine-granularity state identification and representative sequence excavation for application in energy management strategy

Developing scenario driving cycles (Sce-DCs) is significant for calibrating energy management strategies (EMSs), especially for in-plant specialized vehicles. However, existing methods always use micro-trip segments as basic units for state identification, which neutralize DCs' salient features due to their coarse granularity. Additionally, historical state transition sequences (STSs), which remarkably affect time-series characteristics of DCs are usually neglected. Therefore, a multi-pattern fine-granularity state identification method is first proposed to obtain fine-granularity segments. By these fine-granularity segments, salient features can be distinguished and states of DCs can be identified. With the identified states, an improved Prefixspan algorithm is proposed to mine representative STSs. The historical STSs are fully considered by the algorithm's sequence iteration characteristic, culminating in the utilization of integer programming to solve the number of each excavated STS for synthesizing the Sce-DC. Finally, corresponding DC segments are extracted and synthesized to form the Sce-DC. The G-test confirms the consistency of time-series characteristics of the proposed Sce-DC with the database, and the representativeness of kinematic features is improved by 34.07 % versus Markov-Monte-Carlo. Compared to standard DCs, the calibrated EMS of the fuel cell electric vehicle based on the proposed Sce-DC reduces the deviation from the theoretical optimum by up to 9.42 %.