Use of neural network and multivariate statistics in the assessment of pellets produced from the exploitation of agro-industrial residues.

Abstract

The production of pellets from residual biomass generated monocropping by Brazilian agribusiness is an environmentally and economically interesting alternative in view of the growing demand for clean, low-cost, and efficient energy. In this way, pellets were produced with sugarcane bagasse and coffee processing residues, in different proportions with charcoal fines, aiming to improve the energy properties and add value to the residual biomass. The pellets had their properties compared to the commercial quality standard. Artificial neural networks and multivariate statistical models were used to validate the best treatments for biofuel production. The obtained pellets presented the minimum characteristics required by DIN EN 14961-6. However, the sugarcane bagasse biomass distinguished itself for use in energy pellets, more specifically, the treatment with 20% of fine charcoal because of its higher net calorific value (17.85MJ·kg-1) and energy density (13.30 GJ·m-3), achieving the characteristics required for type A pellets in commercial standards. The statistical techniques were efficient and grouped the treatments with similar properties, as well as validated the sugarcane biomass mixed with charcoal fines for pellet production. Thus, these results demonstrate that waste charcoal fines mixed with agro-industrial biomass have great potential to integrate the production chain for energy generation.