Improving Li-ion battery parameter estimation by global optimal experiment design

Li-ion batteries are a key enabling technology for electric vehicles and determining their properties precisely is an essential step in improving utilization and performance. Batteries are highly complex electrochemical systems, with processes occurring in parallel on many time- and length-scales. Models describing these mechanisms require extensive parametrization efforts, conventionally using a combination of ex-situ characterization and systems identification. We present a methodology that algorithmically designs current input signals to optimize parameter identifiability from voltage measurements. Our approach uses global sensitivity analysis based on the generalized polynomial chaos expansion to map the entire parameter uncertainty space, relying on minimal prior knowledge of the system. Parameter specific optimal experiments are designed to maximize sensitivity and simultaneously minimize interactions and unwanted contributions by other parameters. Experiments are defined using only three design variables making our approach computationally efficient. The methodology is demonstrated using the Doyle-Fuller-Newman battery model for eight parameters of a 2.6 Ah 18,650 cell. Validation confirms that the proposed approach significantly improves model performance and parameter accuracy, while lowering experimental burden.