Influence of the Forgetting Factor in the Recursive Least Squares RLS Algorithm on the Quality and Precision of the Identified Parameters in a DC Corona Discharge

In this work, we present a contribution on the applicability of the recursive least squares method used for the parametric identification of a corona discharge phenomenon at small distances. Furthermore, we show the influence of the choice of the forgetting factor for a better performance of the identification operation and the quality of estimation of the identified parameters. The identification process is based on experimental input/output measurements. The validation of the parameter results is done by a physical analysis of the behaviors of these parameters and by comparing the output calculated according to these parameters, with the real output obtained experimentally. The results show that with a constant forgetting factor close to 1 ( $\lambda = 0.99$ ), parameter quality improves but output accuracy may vary. In contrast, a variable forgetting factor enhances both parameter quality and model output consistently. A good agreement observed between the real and calculated outputs confirms both the good choice of the forgetting factor and the precision of the estimated parameters as well as the validity of the identified model in general.