MUNICIPAL WASTE HEATING VALUE MODELLING USING COMPUTATIONAL AND MATHEMATICAL TECHNIQUES

Waste-to-Energy (WtE) processes are widely used to convert solid waste into energy. A WtE plant can contribute to diversify the energy mix and increment the energy supply security of a country. Predict the amount of energy - the heating value - that can be obtained from the waste used as fuel is necessary for the proper operation of a WtE plant. Work with MSW, both in the landfill as in the laboratory, is an unhealthy activity, and requires skilled labour and specialized equipment, usually with a high cost, which cannot be afforded by undeveloped countries. Usually, waste heating value forecasting models are used to minimize the experimental work and, in this sense, it is important to know which techniques can produce good models with a minimum amount of samples. In this paper, three different techniques were compared: Multivariate Linear Regression (MLR), MultilayerPerceptron Artificial Neural Network (ANN), andExtreme Learning Machine (ELM). 36 data samples were available; 26 samples were used as modelling data and 10 samples were used as validate data. As a result, both MLR and ELM techniques had good indexes, with a slight advantage to the ELM, with aMAPE of 4.84% and a MSEof 0.965.