Virtual output estimation in a bioreactor using a generalized super-twisting algorithm

This paper presents the application of a generalized super-twisting algorithm to estimate a critical variable needed for implementing real time optimization feedback controllers in a certain class of bioreactors. A bank of redundant super-twisting observers designed for equivalent systems under a state transformation is used. The procedure is motivated by the fed-batch cultivation of a strain of E. coli that has overflow metabolism, where a useful variable to estimate is the difference between two reaction rates. The results show the applicability of the procedure and illustrates some compromises between the accuracy and the measurement noise.