Neural Modelling of a Yeast Fermentation Process Using Extreme Learning Machines
This work details development of dynamic neural models of a yeast fermentation chemical reactor using Extreme Learning Machines (ELM). The ELM approach calculates very efficiently, without nonlinear optimisation, dynamic models, but only in the non-recurrent serial-parallel configuration. It is shown that in the case of the considered benchmark the ELM technique gives models which are also quite good recurrent long-range predictors, they work in the parallel configuration (simulation mode). Furthermore, properties of neural models obtained by the ELM and classical (optimisation-based) approaches are compared.