Modelling of biohydrogen generation in microbial electrolysis cells (MECs) using a committee of artificial neural networks (ANNs)

Abstract

The enhancement of hydrogen yield in microbial electrolysis cells (MECs) requires a robust process model that accurately relates the effect of anodic physicochemical input variables to the process output. Artificial neural networks (ANNs) have been used for the modelling of complex and non-linear processes. This paper reports the modelling of biohydrogen yield in MECs by using a committee of five ANNs. A topology of 6–(6, 8, 11, 12, 14)–1 was adopted, corresponding to the number of neurons of inputs, hidden (varied) and output layers. The ANN inputs were substrate type, substrate concentration, pH, temperature, applied voltage and reactor configuration. Model development was carried out with 50 data points from 15 published studies. The coefficients of determination (R2) between the experimental and predicted hydrogen yields for the five models were as follows: 0.90, 0.81, 0.85, 0.70 and 0.80. Model validation on new MEC processes showed a strong correlation between the observed and predicted hydrogen yields. Sensitivity analysis revealed that the performance of MEC was highly affected by variations in the substrate type, followed by applied voltage, substrate concentration, pH, MEC configuration and temperature in decreasing order. This study showed that the committee model accurately modelled the non-linear relationship between the considered physicochemical parameters of MEC and hydrogen yield, and thus could be used to navigate the optimization window in MEC scale-up processes.