Microbial fuel cells (MFCs) are bioelectrochemical devices which use micro-organisms as catalyst for electrogenesis at the anode; oxidizing biodegradable substrate to produce electrical current. MFC power output is a function of many factors; including pH, temperature, loading rate, flow rate and electrical load. The study presents a system identification approach to determine a set of linear dynamic black box models able to quantify and represent specific nonlinear characteristics of a MFC. A sandwich-type MFC was subjected to varying electrical loads of various pseudo-random and step inputs, while observing the MFC voltage. Nonlinear behaviour was inferred from assumed piecewise linearised first order dynamic responses, at different operating points. The time constants increased from 0.5 s with PRBS loading of 100–150 Ω, to 6.2 s at 950–1 kΩ; although steady state gain varied little, (0.12–0.20 mV Ω−1). This suggests that the MFC's non-linear behaviour, dependent on operating conditions, may be adequately represented by a series of linear models. System identification suggested that linear 4th order ARX models produce the best fit. However, reasonable prediction was observed using piecewise linearised first order models. The models could be used to design and optimize controllers to regulate power and/or voltage generation.