The aim of this study is to develop a modeling approach able to handle microbial diversity both in normal and abnormal situations. Normal situations are defined as changing input characteristics that do not imply process imbalance while abnormal situations are illustrated by the presence of toxicant into the reactor. In both cases, modeling is performed by adding a stochastic term on top of a well defined and well structured model such as the IWA Anaerobic Digestion Model No1. Experimental data from a 1 m3 pilot scale anaerobic digester treating wine distillery wastewater are provided to demonstrate the applicability of this approach. Discussion also highlights that monitoring of digesters might receive a renewed consideration in the near future in order to handle explicitly microbial diversity within the control objectives.