Optimal control strategies of lithium-ion battery based on electrochemical thermal life model for hybrid electric vehicle and fast charging applications

Abstract

Lithium-ion batteries (LiBs) are the most crucial component of electric vehicles (EVs) and hybrid electric vehicles (HEVs) with respect to cost and performance. Development of optimal control strategies for EVs and HEVs is essential for safe and durable battery use. For analysis and controller design purposes, reduced-order electrochemical-thermal-life models (ROM) are developed, validated, and optimized considering accuracy and computational time. By utilizing the model, optimal energy management strategies (EMSs) are developed for HEVs, and optimal Fast Charging (FC) protocols are developed for EVs. HEVs have complex configurations using multiple power sources, including engines and single or multiple motors. EMS is a control strategy that decides the power split between the power sources. However, battery degradation has not been considered, which is necessary for durable battery usage. Therefore, a new EMS is developed that improves fuel efficiency (FE) and suppresses the degradation of the battery. A hybridized two layer algorithm that combines multi-objective nonlinear model predictive control (NMPC) with a rule-based (RB) algorithm is proposed as a new EMS that is called RB-NMPC. The RB-NMPC is designed to optimize the torque split between the engine and electric motors while maintaining the maximum and minimum constraints of each component. The proposed EMS is incorporated into control-oriented vehicle models, and their performances are analyzed for different driving cycles by comparing with RB, dynamic programming (DP), and NMPC. The long charging time of EVs is a remaining issue for their further commercialization. Based on the validated ROM, two new FC protocols have been developed considering thermal and aging effects. Firstly, we propose an optimized MCC (O-MCC) charging protocol suppressing Lithium Plating (LiP) based on the battery’s State of Health (SOH). Secondly, we propose an optimized MCC protocol with negative pulses (O-MCC+NP) to simultaneously suppress LiP formation and recover lithium-ions via lithium stripping (LiS). The current amplitudes are optimized using NMPC algorithms under constrained LiP. Pulse frequency is determined experimentally, reducing heat generation associated with diffusion resistance by Distribution of Relaxation Time (DRT) analysis. The proposed two new charging protocols are experimentally tested and compared with commercial charging protocols. Lastly, accurate and real-time estimation of SOH is also important for the safe and durable use of HEVs and EVs. Machine Learning (ML) based data-driven approaches gaining popularity in industries, offering a promising alternative to model-based state estimation methods without any knowledge of electrochemical principles. However, their accuracy for FC conditions has not been verified by using only real-time obtainable data. Therefore, we propose an optimized Neural Network (NN) model that achieves accurate SOH predictions across various temperature ranges and charging profiles, which is essential for real-time applications in HEVs and EVs. Moreover, the ML-based SOH estimation algorithm is incorporated with the O-MCC protocol, verifying the performance in Battery-in-the-loop (BIL) system. The model improvement considering mechanical degradation and verifying the control algorithms for batteries with different chemistries are considered as future works.