Abstract
This study investigates the presence of heteroscedasticity in supercapacitors under nonlinear operating conditions commonly encountered in automotive applications. We compare various robust techniques for parameter identification and propose an identification technique that does not increase in complexity with the number of parameters nor require a tuning procedure, called Least Squares with Orthogonal Distances (LSOD). Experimental data from two distinct types of supercapacitors were utilized, and multiple algorithms were employed to test for the presence of heteroscedasticity. Our results indicate that heteroscedasticity is pervasive across all tested conditions, necessitating robust identification techniques to manage non-constant error variability. The LSOD method demonstrated robustness and stability in most conditions, although high current conditions posed significant challenges to all algorithms evaluated. The study underscores the criticality of operating conditions in the behavior of supercapacitors, with high-current conditions being more prone to heteroscedasticity and leading to unstable or minimum-phase behavior.
