Optimizing Electric Vehicles efficiency with hybrid energy storage: Comparative analysis of rule-based and neural network power management systems

The current energy storage solutions for electric vehicles (EVs), powered by a single source such as batteries, fuel cells, flywheels, or supercapacitors (SCs), hinder efforts to make EVs comparable to conventional vehicles due to limitations in energy density, power density, and cycle life inherent in each source technology. The hybridization of different energy sources represents a significant advancement in addressing the efficiency and reliability challenges of EVs, potentially surpassing single-source systems. However, this hybridization adds complexity to managing diverse energy sources within the power management system (PMS). Our study introduces a dual-energy storage solution for EVs, combining a lithium-ion battery as the primary power source with a SC as the auxiliary energy reserve. We conduct a comparative analysis of several PMS methods, including rule-based approaches (filtering, limitation, dynamic battery power limiting (DBPL) and fuzzy logic) and neural network (NN). Through detailed simulations and experimental implementation using dSPACE 1140 and MATLAB, we assess these methods by evaluating battery stress reduction, cost-effectiveness, computational requirements, and the SC charging level at the driving cycle end. Obtained results show that some basic rule-based methods can degrade battery life even with the addition of SCs, compared to EVs with a single battery source. Enhanced rule-based methods such as Fuzzy Logic Control (FLC) and DBPL show significant potential, comparable to machine learning (ML) techniques. The conducted comparative analysis reveals that DBPL and NN methods are crucial for extending Li-ion battery life, maintaining optimal SC charging, and maximizing braking energy recovery.