Diseño de una red neuronal artificial para la predicción del poder calorífico superior a partir del análisis termogravimétrico (tga) de residuos de naranja

Abstract

The objective of this study was to develop an artificial neural network to predict the higher heating value in the orange peel residues in the Caluma canton, to create a database with 110 data. To develop the prediction model, experimental data from thermogravimetric analysis and proximal analyses were used. The network design was carried out in Matlab and Simulink Student Suite using four input variables, 10 neurons in the hidden layer, one output variable and the Levenberg Marquardt algorithm. Four input variables were selected for the neural network model, these are the percentage of humidity, fixed carbon, volatile matter, and ash, and as an output variable the higher calorific value. A statistical P-value of 0.999 allowed the prediction model to be validated with 95% confidence, confirming that there is no significant difference between the real values of the higher heating value and those predicted by the artificial neural network. The results indicate that the artificial neural network proves to be efficient in predicting the higher heating value in the orange peel residues analysed by thermogravimetry, using a heating rate of 15 ° C / min in an inert atmosphere with nitrogen injection at 20 mL / min. It is recommended to use the neural network in projects related to the pyrolysis or gasification of biomass that uses orange peel residues as a source of bioenergy.