Battery Identification Methods Based On Equivalent Circuit Model

Development of an intelligent battery diagnostic system is a necessity for future transportation industry. These technologies will have the potential to create profound impact in other industries such as portable electronics. This thesis reports on battery identification methods that are primarily engineered to detect the chemistry, number of cells, and state of charge in an unknown package of batteries. The proposed methods have the potential to be used for condition monitoring in a known set of batteries thereby, creating a health monitoring apparatus that can be an integral part of a battery management system using any of the prominent lead acid, lithium-ion, and Nickel Metal Hydride batteries. The proposed methods are based on distinct signatures that one can identify in a relatively straightforward equivalent circuit of a battery. These signatures are extracted using time domain diagnostics and are used in combination with nonlinear mappings such as exponential regression and artificial neural networks for pattern recognition purposesThis thesis presents the design and development of three battery identification methods based on a single RC equivalent circuit model. The first method compares measured circuit parameters with lookup tables using MSE analysis to identify chemistry, cell count, and SOC of the battery. The second method uses an artificial neural network to identify battery chemistry based on measured circuit parameters. The final method uses an artificial neural network to identify battery chemistry and SOC based on raw voltage waveforms, bypassing the need to calculate equivalent circuit parameters.