Stochastic optimal control for series hybrid electric vehicles

Abstract

Increasing demand for improving fuel economy and reducing emissions has stimulated significant research and investment in hybrid propulsion systems. In this paper, we address the problem of optimizing online the supervisory control in a series hybrid configuration by modeling its operation as a controlled Markov chain using the average cost criterion. We treat the stochastic optimal control problem as a dual constrained optimization problem. We show that the control policy that yields higher probability distribution to the states with low cost and lower probability distribution to the states with high cost is an optimal control policy, defined as an equilibrium control policy. We demonstrate the effectiveness of the efficiency of the proposed controller in a series hybrid configuration and compare it with a thermostat-type controller.

Concurrent Optimization of Vehicle Dynamics and Powertrain Operation Using Connectivity and Automation

“…The power management control algorithm in HEVs and PHEVs determines how to split the power demanded by the driver between the thermal and electrical subsystems so that maximum fuel economy and minimum pollutant emissions can be achieved. Developing the control algorithm in HEVs and PHEVs constitutes a challenging control problem and has been the object of intense study for the last 20 years [23][24][25][26]. A significant amount of work has been proposed on optimizing the power management control in HEVs using dynamic programming (DP) [27].…”

Supervisory Power Management Control Algorithms for Hybrid Electric Vehicles: A Survey

“…Although the effectiveness of the power management control algorithm was validated through simulation in a series hydraulic hybrid vehicle, the algorithm could probably be also successfully applied in a series HEV. Another proposed online control approach addressed this problem by modeling HEV operation as a controlled Markov chain [118]. The stochastic optimal control problem was treated as a dual constrained optimization problem using the average cost criterion.…”

Energy and power management in a series Hybrid Electric Vehicle using Selective Evolutionary Generation

“…Solution techniques for the SHEV energy and power management problem take a number of forms, including rule-based control strategies [6], [7], dynamic programming [4], [8], [9], Pontryagin's maximum principle [10], [11], equivalent consumption minimization strategies [11], [12], convex optimization [13], support vector machines [14], smoothed transitions between steady-state optimal operating points [15]- [17], online control [16], [18], Model Predictive Control (MPC) [17], stochastic MPC [19], [20], stochastic optimal control [21], etc.…”

“…SHEV operation is formulated similar to [17] to facilitate a similar future investigation into overall powertrain efficiency improvement, an emphasis that differs from previously-published on-line approaches [16] and [18] that optimize fuel consumption. The present work is also motivated, in part, by [21], which models SHEV operation as a controlled Markov chain using the average cost criterion to show that a stochastic control policy that yields higher probabilities of SHEV states with low cost and lower probabilities of SHEV states with high cost is an optimal control policy.…”