Dynamic learning rate (ηD) for recurrent high order neural observer (RHONO): Anaerobic process application

In this paper, a dynamic learning rate, for recurrent high order neural observer (RHONO), is proposed. The dynamic learning rate depends on the pH on-line measurement. The main objective is to improve learning of the neuronal network in presence of disturbances, which is obtained by increasing the performance of the neuronal observer by means of the dynamic learning rate. The learning algorithm is based on an extended Kalman filter. The applicability of the proposed dynamic rate is illustrated via simulation, as applied to a RHONO for an anaerobic process.