Lithium-ion battery responses to bipolar pulse perturbations of less than two minute duration and one C-rate amplitude are studied as general-purpose diagnostics signals that encode the cell impedance, remaining charge, and degradation level. It is shown that the information is derived from a combination of the linear and nonlinear system dynamics of the electrochemical overpotentials, open-circuit voltage change, and hysteresis of the cell.
Experimental data is analyzed using an equivalent circuit composed of a conventional resistor-capacitor pair model, a square-root-order convolution-defined diffusion element, and a piece-wise-linear open-circuit voltage element. This bipolar pulse model disaggregates the battery voltage response into its constituent dynamics and allows the nonlinearities to be isolated. The nonlinearities are crucial features which allow the battery charge, health, and incremental capacity features to be regressed directly from the pulse voltage response using ridge regression and feedforward neural networks. Assessment of different pulse shapes suggests that the diagnostics power of the pulse may increase with higher amplitude and shorter duration. Real-world applications are then investigated, including the estimation of charge imbalance using the series-module pulse response, and state-space formulation of the convolution-defined diffusion element.
Further refinement of the pulse techniques could simplify battery diagnostics by providing, from a single pulse diagnostic, the key states of charge, health, and power necessary to operate a reliable system.
| Identifer | oai:union.ndltd.org:columbia.edu/oai:academiccommons.columbia.edu:10.7916/ttrn-5110 |
| Date | January 2024 |
| Creators | Li, Alan Gen |
| Source Sets | Columbia University |
| Language | English |
| Detected Language | English |
| Type | Theses |
Page generated in 0.0085 seconds