DATA IDENTIFICATION OF BIOMASS FOR ENERGY USING NEURAL NETWORKS APPLIED TO THERMOGRAVIMETRIC CURVES

This work evaluates the use of TGA curve markers (similar to genetic markers) to obtain information on biomass. That is, multiple values of percentage of residual weight with respect to the initial one at specific temperatures of the TGA curve are used to identify specie and wood and leaf mixtures.

The data obtained from the TGA analyzes for different mixtures of leaves and wood (100% leaf, 50% leaf and 50% wood and 100% wood) of 7 species are presented (poplar (Populus sp.), caper (Euphorbia laurifolia), alder (Alnus acuminata), arupo (Chionanthus pubescens), cypress (Cupressus macrocarpa), eucalyptus (Eucalyptus globulus), linden (Sambucus nigra L.) and pine (Pinus radiata))

From each of the thermogravimetric curves, the residual weights of the sample at fixed temperatures have been selected (100ºC, 175ºC, 200ºC, 325ºC, 400ºC, 475ºC and 550ºC). These weights represent the input to a neural network to identify both the species and the percentage of leaves and wood.

The last sheet of the file shows the assignment of values to each species and the final combination of the mixture of leaves and wood, output of the neural network.