Multi-layer perceptron artificial neural network (MLP-ANN) prediction of biomass higher heating value (HHV) using combined biomass proximate and ultimate analysis data

Biomass is an important renewable energy source and studies on its properties help determine the suitable choice of process to harness this energy for a specific biomass type. Higher heating value (HHV) is a measure of the biomass energy content, and recent studies have now focused on non-empirical techniques of predicting it. The aim of this study was to utilise multi-layer perceptron artificial neural network (MLP-ANN) modelling to predict the higher heating values of biomass based on 210 lines of combined proximate and ultimate analysis data. The combined data were more suitable for the modelling with ANN architecture of 8-10-1 compared to the singularised proximate and ultimate analysis results considering the R2 and error values for the training, testing and validation sets. At the testing phase, the MSE, RMSE and R2 obtained when using the combined data was 0.8097, 0.8998 and 0.9249, respectively; the high R2 for both the testing and validation sets show good estimation and the generalisation capacity of the ANN model developed. Model validation via parity plot and direct comparison of results revealed that the model does not show any prediction bias and can capture even outliers. Also, comparison with other models revealed that the current study represents an improvement on the status quo. The findings of the study afford a platform where biomass HHV can be obtained faster, at a lower cost and can be determined in real time.