An Extended Range Electric Vehicle Backward-looking Model Accounting for Powertrain Transient Effects 2020-01-1442
Extended range electric vehicles (EREVs), including Plug-in Hybrid Electric Vehicles, are taking an increasing share of the current automobile market, because they offer the functionality of pure electric driving while successfully tackling the major weaknesses of battery electric vehicles (BEVs) such as relatively high price, short range, and long charging duration. Since the EREV powertrain structure is based on different power sources, a key vehicle design activity is related to development of an optimal control strategy for achieving a high fuel economy potential. The central role in developing an optimized energy management strategy is related to availability of computationally-efficient, high-fidelity EREV powertrain model. The state-of-the art modeling techniques are related to two basic approaches: (i) a backward-looking model whose only state variable is battery state-of-charge (SoC), and (ii) a forward-looking model that takes into account the powertrain dynamics effects such as inertia, halfshaft compliance, engine/e-motor/actuator dynamics, and low-level control system dynamics. The former is computationally very efficient and is typically used for control trajectory optimization (usually by using dynamic programming) and initial design of supervisory energy management control strategy. The latter is used for fine-tuning of the overall control strategy and final simulation verifications. Having in mind the above, there is clearly a necessity to design a backward-looking model which would include transient effects to approach the accuracy of the forward-looking model. This paper proposes such a model, where the transient effects are accounted for through additional fuel consumption increment and battery state-of-charge decrement in each simulation step. These effects are modelled by employing data-driven regression techniques, where data have been collected through numerous simulations of the full, forward-looking model. The proposed, extended backward model is validated against the forward model for the case of Chevrolet Volt EREV powertrain, in order to illustrate its performance in terms of accuracy (close to forward model) and computational efficiency (close to backward model).
Jure Soldo, Branimir Skugor, Josko Deur