A novel LSTM-based EMS and interleaved DC-DC boost converter topology for real-time driving conditions in HEVs

The goal is to develop an adaptive energy management method that can predict driving circumstances, reducing stress on the primary storage battery in a hybrid electric vehicle (HEV). To improve battery lifespan and overall efficiency, power allocation among multiple energy storage sources (ESS) must go beyond traditional rule-based schemes. This paper offers a promising solution to address these challenges. Standardized driving cycles, such as the Urban Dynamometer Driving Schedule (UDDS), are used as load torque to evaluate the power demand in hybrid electric vehicles (HEV). First, feature extraction techniques improve the extraction of important features from historical driving cycle data, thereby facilitating the long short-term memory (LSTM) based on multistep velocity predictors using a warm-up algorithm. This predicted future velocity profile serves as input to the HEV system. Second, within the hybrid electric vehicle (HEV) system, the battery is linked to a modified interleaved DC-DC boost converter. The results show that this converter affects factors such as ripple, converter loss of 29%, state of charge (SOC) level, and efficiency improvement to 98%. Finally, the energy management system (EMS) with the modified rule-based scheme is implemented with neural networks for optimization. This EMS is responsible for continuously adjusting the power allocation. Simulation results show a significant average reduction of about 33.33% at peak battery level, thereby extending battery life. Compared with conventional rule-based EMS methods, the proposed strategy significantly improves efficiency by reducing the total energy loss by about 10.36%. These results emphasize the reliability and robustness of the proposed strategy.