Artificial neural network model for the prediction of kinetic parameters of biomass pyrolysis from its constituents

- Artificial neural network models were developed to predict biomass pyrolysis.
- A total of 150 data from different sources were considered.
- The relative importance of each biomass constituent was determined.
- Contour plots were generated to predict the kinetic values from biomass composition.

This study applied artificial neural networks (ANN) for constructing the correlation between biomass constituents and the kinetic parameters (activation energy (Ea), pre-exponential factor (k0) and reaction order (n)) of biomass pyrolysis. Three ANN models were developed, one for each of the three kinetic parameters. A total of 150 experimental thermogravimetric analyses from a diverse range of biomass compositions were used to develop and test the networks. The relationships between the main biomass components and the output parameters were non-linear and could potentially be predicted by the selected ANN models (R2 > 0.9). Using a mean standard error limit of 0.001, the number of neurons in the hidden and the output layer and the model parameter weights and biases were optimized, with 20, 17 and 30 neurons, for log k0, log Ea and log n, respectively. The generated contour plots revealed that cellulose required the highest k0, Ea and n values, as well as the non-linearity and complexity of the system.

155