Optimal control of a parallel hybrid electric vehicle with traffic preview

© 2011 Dr. Tae Soo KimHybrid electric vehicles (HEV) offer lower fuel consumption relative to conventional powertrains through the added functionality provided by an electric motor. The improvement in fuel consumption achievable through hybridisation is heavily dependent on scheduling the motor and engine appropriately, which ideally requires feed-forward information about the conditions to be encountered by the vehicle. Telemetry systems can be used to estimate future traffic conditions, and consequently a suitably equipped vehicle may control its own velocity and torque split ratio to maximise fuel efficiency under certain operating constraints. This thesis develops a real-time implementable controller for torque split and vehicle velocity of an HEV with telemetry, which alleviates computational issues of a multivariable optimal control problem by utilising optimal control theory. Using continuously updated traffic information provided by telemetry in a finite moving horizon, three controllers are proposed and analysed. The implementation in a receding horizon framework satisfies traffic constraints while sustaining the battery state of charge to within acceptable margins. First, a numerical optimisation technique is appointed in solving the multivariable optimisation problem and its computational limitations are identified especially with long preview length. In response, an optimal controller for the two control inputs is developed analytically using Pontryagin’s Minimum Principle. To handle the discontinuities in the vehicle dynamics and switching objective function between propelling and braking transition, two-stage optimality conditions have been adopted. Closed form solutions obtained are applied in receding horizon at minimal computational effort. The validity of common modelling assumptions that simplify the controller is investigated under realistic driving simulations over a range of traffic preview lengths. Simulations showed that the fuel consumption obtained using the proposed controller is within 1.2% of the benchmark result obtained by non-real time DP-based optimisation. Further extending the analysis, common simplifying assumption on engine efficiency in the controller development is relaxed to obtain more generalised controller which can be applied to a class of vehicles whose engine efficiencies are approximated in a quadratic function of engine torque. Portability of the controller is demonstrated on three different vehicle models. In addition, the impact on the proposed controller of preview duration and measurement noise is investigated. It is demonstrated that there is a maximum length of traffic preview at which powertrain hybridisation will deliver fuel economy benefits. The choice of appropriate preview length depending on the type of telemetry error present is justified