Sensitivity analysis for reliable state-of-health estimation based on battery partial charging

Lithium-ion batteries are essential for powering modern technologies like portable electronics and electric vehicles because of their energy density and efficiency. However, batteries incrementally degrade, leading to a reduction in their health. Therefore, accurately predicting their state of health is vital for optimal operation, reducing cost, enhancing safety, and promoting sustainability. Traditional methods, even though they are accurate, remain impractical as they often rely on complete charge-discharge cycles, while real-world operation involves dynamic discharging and partial charging. We compare four estimation methods, utilizing partial charges from realistic battery operation, based on five batteries subject to varying loads and temperatures. The main objective is to analyze the sensitivity of the information used to build these models. Our analysis shows that we can build models with errors below 0.5% using information from a few random partial charges. That is, reliable models can be built using limited inconsistent data, thus improving economic feasibility.