Improving Fuel Economy of Thermostatic Control for a Series Plugin-Hybrid Electric Vehicle Using Driver Prediction

This study investigates using driver prediction to anticipate energy usage over a 160-meter look-ahead distance for a series, plug-in, hybrid-electric vehicle to improve conventional thermostatic powertrain control. Driver prediction algorithms utilize a hidden Markov model to predict route and a regression tree to predict speed over the route. Anticipated energy consumption is calculated by integrating force vectors over the look-ahead distance using the predicted incline slope and vehicle speed. Thermostatic powertrain control is improved by supplementing energy produced by the series generator with regenerative braking during events where anticipated energy consumption is negative, typically associated with declines or decelerations. Matlab and Simulink (Colgren, 2007) are used both to run the prediction and optimization algorithms as well as simulate a rear-wheel drive, series plug-in, hybrid-electric vehicle, a city road-network, and multiple simulated vehicle drivers each with different driving characteristics. Initial results show an improvement of 8.4% in petroleum energy consumption over a non-predictive thermostatic propulsion controller.