Fuel Cell Hybrid Electric Vehicle Control: Driving Pattern Recognition Techniques to Improve Vehicle Energy Efficiency

Abstract

<div class="section abstract"><div class="htmlview paragraph">Hydrogen technologies have been widely recognized as effective means to reduce Greenhouse Gases emissions, a crucial issue to target a Carbon-free world aimed by the European Green Deal. Within the road transport sector, electric vehicles with a hybrid powertrain, including battery packs and hydrogen Fuel Cells (FCs), are gaining importance owing to their adaptability to a wide variety of applications, high driving mileages and short refueling times. The control strategy is crucial to achieve a proper management of the energy flows, to maximize energy efficiency and maximize components durability and state of health. This work is focused on the design of an integrated Energy Management Strategy (EMS), whose aim is to minimize the hydrogen consumption, by operating the FC mainly in the high efficiency region while the battery pack works according to a charge sustaining mode. The proposed EMS is composed of a control algorithm and a supervisor. A series of fuzzy controllers have been implemented: their Membership Functions have been designed by starting from a first guess and subsequently they have been trained through a Genetic Algorithm, targeting the optimal results previously obtained by a Dynamic Programming approach on specific driving cycles, resulting from a k-means clustering algorithm. On the other hand, within the supervisor, a Driving Pattern Recognition algorithm has been implemented, able to detect in real-time the actual driving conditions and to switch adaptively between the proper sub-optimized fuzzy controller options. The analysis has been performed for a microcar application, with four 2kW-nominal in-wheel motors, two 2kW rated power FCs and a 5.1kWh-capacity battery pack. The FC model has been validated through experimental tests. Results show that the system is able to manage the battery State of Charge around the target value (70%), considering two driving cycles, and to maintain the sub-optimal performances with an increase in hydrogen consumption of only 3.7 % if compared to the global optimum of Dynamic Programming results.</div></div>