(15b) Analysis of Lithium-Ion Battery Management System and Degradation Mechanism

Battery technologies, particularly lithium-ion batteries (LIB) have shown promise in automotive applications. To ensure reliable, and safe operation, a battery management system (BMS) is used, which acts based on the battery model. In past, equivalent circuit models (ECM) were employed, estimating simple battery state and parameters (e.g. ohmic resistance, open circuit voltage) using fixed safety constraints (e.g. fixed cut-off and cut-in voltage). This rendered simpler computation and easy implementation. But these models have limitations in providing insights into the internal behavior of the batteries (e.g. internal temperature, state of charge). Knowledge about electrochemical behavior is important because of two major reasons - as the battery ages, the operability regions changes (irreversible degradation due to loss of ions, active material, or electrolyte)., thus the fixed safety constraints may not guarantee safer and optimal operation. Secondly, newer automotive technologies demand broader operating conditions (e.g. fast charging, wider temperature profiles, etc.) which calls for advanced algorithms.

Electrochemical models (EM) play a major role in the future BMS. EM’s are based on physiochemical mechanisms, resulting in a more accurate prediction while describing the internal states of the battery at wider operating ranges. However, the governing equations are non-linear and have multiple partial differential equations (PDE) that make them computationally expensive. Moreover, a large number of physical parameters are required which are often dependent on composition and temperature (e.g. thermodynamic potential, ionic conductivity, diffusion coefficient, transference number). Estimating these parameters with achievable accuracy continues to be a challenging job.

This study addresses two pivotal questions – how can we reduce or transform the complex full order EM maintaining the same accuracy at a wider operating window? We explore the suitability of different methods to convert PDE’s (e.g. Finite Element Method, Pade approximation, Orthogonal decomposition, etc.) into simpler equations. Secondly, how to systematically estimate parameters without information loss? Parameter estimation techniques (e.g. Fisher Information Matrix, Particle Swarm Optimization, Genetic Algorithm) are studied to characterize the battery degradation and shed light on the cause-effect mechanisms.