The nonlinear phenomenon of the profile of substrate concentration and hydrogen production rate over 16 retention days in a 4 L double chamber of a microbial electrolysis cell (MEC) for bioelectrochemical production of hydrogen from sago wastewater validates the mathematical modeling results based on simplified microbial biofilm growth. The stoichiometric reaction and kinetics affect the substrate concentration curve behaviour, but the effects also include the bioelectrochemical balance for hydrogen production rate. The artificial neural network (ANN) predicts the experimental hydrogen production rate according to the input of pH of the catholyte at controlled applied potential of 0.8 V and current density of 0.632 A‧m-2. The convex method assists the model in finding the optimal input values that lead to the minimum mean square error (MSE) between modelling and experimental data. Evaluation of the COD removed efficiency, coulombic efficiency, and energy efficiency determines the process limit of the model MEC. At an optimum applied potential of 0.485 V, anode surface area of 0.098 m2, anodic chamber volume of 4 L, and initial substrate concentration of 2,500.99 mg‧L-1, the MEC model reached maximum steady-state percentage at 81.99% of COD removed efficiency, 69.01% of Coulombic efficiency, and 7.47% of energy efficiency.