Ribera, Mattia (2025) Model-Based and Data-Driven SoH Estimation for Lithium-Ion Batteries: A Comparative Study with Deployment Guidelines. [Tesi di dottorato]

Documento PDF Ribera_Mattia_38.pdf Visibile a [TBR] Amministratori dell'archivio Download (36MB) | Richiedi una copia

Abstract

Lithium-ion batteries enable key use cases on two fronts: in e-mobility they power vehicles and support smart-charging infrastructure; in stationary applications they shave peaks, provide flexibility services, and enhance grid resilience. When coupled with variable renewables (photovoltaic and wind), storage acts as an energy buffer: it absorbs surplus generation, delivers power during shortfalls, mitigates intermittency, and shifts energy over time. In this setting, accurate State-of-Health (SoH) estimation is essential to ensure availability, safety, and sustainable lifecycle costs. This thesis develops and contrasts two complementary pipelines for SoH estimation using laboratory datasets across multiple lithium-ion chemistries. The first is a model-based pipeline grounded in the Impulse Response (IR) concept and ARMAX modeling. The resulting models are organized into look-up tables (LUTs) indexed by state and operating conditions; the approach offers interpretability, diagnostic capability, and predictable behavior under controlled excitations. The second is a data-driven pipeline that employs neural networks to map preprocesssed voltage–current–temperature sequences to SoH. The thesis details the processing flow and validates the method across different chemistries. Beyond the method-specific conclusions, the work closes with a comparative analysis that highlights key differences not only in final performance, but also in pre-processing requirements, computational cost, and implementation choices.

Downloads

Actions (login required)

Modifica documento