Fast Numerical Powertrain Optimization Strategy for Connected Hybrid Electric Vehicles

Abstract

Connected vehicle (CV) technology allows traffic information sharing that can be utilized to improve fuel benefits of hybrid electric vehicles (HEVs). However, there are challenges from the powertrain optimization perspective. First, optimizing powertrain operations is computational heavy due to the nonlinearities, constraints, and high dimensionality of the problem, especially with long optimization time horizon. Secondly, a slow powertrain optimization means the solution may not be optimal anymore at current time if the traffic states change. In this paper, a fast, near global-optimal, and practical HEV powertrain optimization utilizing vehicle speed prediction in CV environment is systematically formulated and solved. First, the minimum fuel rate and battery state-of-charge rate are calculated for different engine and battery power-split ratios. Then, the nonlinear minimum fuel rate and battery state-of-charge rate are piecewise-linearized by introducing dimensionless variables. The engine operating range, engine dynamics, and battery charge-sustaining constraint are represented by the dimensionless variables and treated as linear constraints. This transformation exploits the problem's structure, known as Separable Programming, which is solved efficiently as a large-dimension constrained linear problem. Fast optimization allows utilization of repeated speed predictions to maintain prediction accuracy. Results show comparable fuel economy with Dynamic Programming with significantly less computation time. Average fuel savings of 4.0% and 10.4% over Pontryagin's Minimum Principle and Rule-Based optimization methods are observed. Experimental results show the feasibility of the optimized engine operating points with fuel benefits over the rule-based method. Simulations with vehicle speed prediction and limited Gaussian uncertainties to emulate CV settings also show satisfactory fuel benefits.