Statistical driver behavior-based power management design with stochastic optimization method for parallel HEVs

Nowadays, predictive control which applies a model to predict the future system behavior is suitable for power management design in parallel HEV. However, both vehicle and driver should be considered together for predicting the system dynamics in the future, especially the driver behavior. In this paper, the driver's action, torque demand, is regarded as stochastic variable which is modelled as Markov process based on known conditioned probability distribution obtained from driver's statistical behaviors. Then, the control maps are obtained by off-line optimization algorithm under consideration of vehicle dynamics and the stochastic future torque demand. With cost function evaluating the equivalent energy consumption, the stochastic optimization problem with chance-constrained is solved by combining scenario approach and vector quantization method. Numerical simulation-based vase studies are demonstrated to validate the proposed design scheme finally.