State-of-Charge and State-of-Health Estimation of Batteries for Urban Air Mobility with Hybrid Propulsion System

다국어 초록 (Multilingual Abstract)

Urban Air Mobility (UAM), enabled by electric vertical take-off and landing (eVTOL) aircraft, is expected to reshape future transportation. To extend flight range and payload, hybrid-electric propulsion (HEP) is adopted, placing the battery under mixed charge-depleting (CD) and charge-sustaining (CS) operation. For safety-critical flight, a Battery Management System (BMS) must estimate State-of-Charge (SOC) and State-of-Health (SOH) reliably across the battery life cycle. However, standardized benchmark profiles for hybrid UAM are limited, and many co-estimation methods—typically tuned on automotive cycles and mild aging—lose robustness in deep degradation as capacity becomes weakly observable and aging-induced model mismatch increases. This dissertation presents a model-based SOC/SOH estimation methodology tailored to hybrid UAM operation. Using Plug-in Hybrid Electric Vehicle (PHEV) profiles as surrogate profiles, the framework is developed and evaluated on experimental data spanning up to 30% capacity fade. Aging data are used to select a parsimonious Enhanced Self-Correcting (ESC) model and the subset of parameters to be estimated online. A Dual Extended Kalman Filter (DEKF) analysis provides practical guidance for covariance tuning and identifies a key failure mechanism in deep degradation: an observability asymmetry between capacity and impedance-related parameters. To address this, a decoupled adaptive framework is proposed: capacity is estimated from charging windows via Weighted Least Squares (WLS), while SOC and impedance parameters are estimated via a DEKF with SOH-triggered baseline parameter updates. Performance is first evaluated on the surrogate profiles under deep degradation, where the proposed framework achieves SOC RMSE < 4.5% and SOH error < 3%. Transferability is then assessed separately using independent datasets spanning different cell types and representative UAM mission profiles, demonstrating the applicability of the estimator to UAM mission conditions and its scalability to a large-format pouch cell within the tested aging ranges.

목차 (Table of Contents)

• 1. INTRODUCTION 10
• 2. BACKGROUND AND LITERATURE REVIEW 14
• 2.1 MISSION STANDARDIZATION CHALLENGES AND SURROGATE VALIDATION FOR HYBRID UAM 14
• 2.2 BATTERY DEGRADATION AND HEALTH METRICS FOR UAM 16
• 2.3 STATE-OF-CHARGE (SOC) ESTIMATION STUDIES 18
• 2.3.1 MODEL-BASED SOC ESTIMATION 18
• 2.3.2 DATA-DRIVEN AND PHYSICS-BASED SOC ESTIMATION 18
• 2.4 STATE-OF-HEALTH (SOH) ESTIMATION STUDIES 19
• 2.4.1 CAPACITY-BASED SOH ESTIMATION 19
• 2.4.2 RESISTANCE- AND POWER-BASED SOH ESTIMATION 20
• 2.4.3 DATA-DRIVEN AND MACHINE-LEARNING SOH ESTIMATION 20
• 2.5 CO-ESTIMATION OF SOC AND SOH 21
• 2.5.1 JOINT ESTIMATION FRAMEWORKS 21
• 2.5.2 DUAL-FILTER AND MULTISCALE FRAMEWORKS 21
• 2.5.3 HYBRID AND DATA-DRIVEN APPROACHES 22
• 2.6 RESEARCH GAPS AND POSITIONING OF THIS DISSERTATION 22
• 3. EXPERIMENTAL SETUP AND DATASET 26
• 3.1 TEST BENCH AND EQUIPMENT 27
• 3.2 TESTING FRAMEWORK 27
• 3.3 DATASET 1: METHODOLOGY DEVELOPMENT (18650-35E) 29
• 3.4 DATASET 2: METHODOLOGY VALIDATION (18650-25R) 33
• 3.5 DATASET 3: METHODOLOGY VALIDATION (P92 POUCH CELL) 38
• 3.6 SUMMARY 41
• 4. BATTERY MODELING 43
• 4.1 EQUIVALENT-CIRCUIT MODEL 44
• 4.2 ENHANCED SELF-CORRECTING (ESC) MODEL 46
• 4.3 PARAMETER IDENTIFICATION METHODOLOGY 47
• 4.4 MODEL COMPARISON RESULTS 50
• 4.4.1 IMPACT OF AGING ON MODEL FIDELITY 52
• 4.4.2 EFFECT OF INCREASING RC PAIRS 52
• 4.4.3 PARAMETER TRENDS AND INTERPRETATIONS 53
• 4.5 FINAL MODEL SELECTION AND DISCUSSION 58
• 5. ESTIMATION METHODOLOGY AND RESULTS 61
• 5.1 DUAL EXTENDED KALMAN FILTER (DEKF) (SCHEME 1) 62
• 5.1.1 MATHEMATICAL FORMULATION 63
• 5.1.2 IMPLEMENTATION FOR ESC-1RC MODEL 66
• 5.1.3 FILTER TUNING 70
• 5.1.4 ESTIMATION RESULTS AND LIMITATIONS 76
• 5.1.5 ROOT CAUSE ANALYSIS: ISOLATING THE INITIAL SOC ERROR 85
• 5.2 DECOUPLED DEKF WITH WEIGHTED LEAST SQUARES (WLS) (SCHEME 2) 90
• 5.2.1 CAPACITY ESTIMATION BY WEIGHTED LEAST SQUARES (WLS) METHOD 90
• 5.2.2 SOC ESTIMATION RESULTS AND LIMITATIONS 92
• 5.2.3 ROOT CAUSE ANALYSIS: ISOLATING THE INITIAL SOC ERROR 96
• 5.3 PARAMETER BASELINE UPDATE STRATEGY (SCHEME 3) 100
• 5.3.1 DETERMINATION OF SOH-TRIGGERED UPDATE INTERVALS 101
• 5.3.2 SOC ESTIMATION RESULTS AND FINAL VALIDATION 104
• 5.4 CONCLUSION 108
• 6. METHOD VALIDATION 111
• 6.1 VALIDATION ON THE DATASET1 112
• 6.1.1 IMPLEMENTATION STRATEGY: SCHEME 3 APPLICATION 112
• 6.1.2 ESTIMATION RESULTS 114
• 6.1.3 ENHANCED VALIDATION: THE IMPACT OF OCV UPDATE 116
• 6.2 VALIDATION ON THE DATASET 3 117
• 6.2.1 IMPLEMENTATION STRATEGY: SCHEME 1 APPLICATION 117
• 6.2.2 ESTIMATION RESULTS 120
• 6.3 SUMMARY 122
• 7. CONCLUSION AND FUTURE WORK 123
• 7.1 SUMMARY AND CONCLUSIONS 123
• 7.2 KEY CONTRIBUTIONS 124
• 7.3 LIMITATIONS AND FUTURE WORK 125
• LIST OF REFERENCE 127