Energy Management Based on Mixed-Integer Nonlinear Model Predictive Control for Hybrid Electric Vehicles

Due to the inherently coupled dynamics of vehicle and powertrain levels, this paper proposes an energy management strategy that co-optimizes the power split and operating mode selection for hybrid electric vehicles. A mixed integer nonlinear model predictive control (MPC) problem is formulated to gu...

Full description

Saved in:
Bibliographic Details
Published in:IEEE transactions on intelligent transportation systems Vol. 25; no. 11; pp. 17432 - 17451
Main Authors: Hou, Shengyan, Chen, Hong, Yin, Hai, Zhao, Jing, Xu, Fuguo, Gao, Jinwu
Format: Journal Article
Language:English
Published: IEEE 01.11.2024
Subjects:
Batteries
battery lifetime
Energy management
Engines
Gaussian process regression
Gears
hybrid electric vehicles
Mechanical power transmission
mixed-integer optimal control problem
Optimization
Real-time systems
ISSN:1524-9050, 1558-0016
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Due to the inherently coupled dynamics of vehicle and powertrain levels, this paper proposes an energy management strategy that co-optimizes the power split and operating mode selection for hybrid electric vehicles. A mixed integer nonlinear model predictive control (MPC) problem is formulated to guarantee the sub-optimality and robustness of the strategy. The optimization objectives are to achieve minimal fuel consumption, battery degradation inhibition and state-of-charge maintenance within the system physical constraints. However, the existence of logic events and continuous variables poses a significant challenge to solve the optimal control problem. A Pontryagin's Minimum Principle (PMP)-Dynamic Programming (DP) based solution method is presented, avoiding segmented linear approximation to the powertrain model and relaxation approach to the problem. The PMP calculates the optimal control sequence and cost for each mode, then determines the optimal operating mode based on DP. Moreover, the Gaussian process regression (GPR) model is developed for vehicle speed prediction to deal with the stochastic uncertainties of driving conditions. Experimental results are demonstrated that the proposed algorithm offers about 2% to 10% cost reduction compared to the conventional MPC, while still keeping relatively close to the result of DP.
ISSN:1524-9050
1558-0016
DOI:10.1109/TITS.2024.3432775