In this work, a variable structure model (VSM) of an anaerobic digestion process was developed. The anaerobic biodegradation process was described by four nonlinear submodels representing methanogenic, chemical oxygen demand overload, acidogenic, and hydrogen-inhibited states of the anaerobic process. At any instant, process dynamics was modeled only by one of the submodels, while the others were considered trailing. The choice of a leading submodel was handled by a knowledge-based system, which analyzed available process variables, such as off-gas composition and reactor pH. The feasibility of the proposed method was demonstrated both by using the VSM to predict the outputs of a comprehensive process model, and the experimental results obtained in a pilot scale anaerobic fixed-bed bioreactor.