Heating value prediction for combustible fraction of municipal solid waste in Semarang using backpropagation neural network

Generalized models to predict the lower heating value (LHV) of municipal solid waste (MSW)

“…The negative or low contribution of food waste to the overall LHV in both regression analysis and ANN illustrates that food waste is not suitable for incineration. Because of the lower carbon content and higher oxygen content in food waste (or organic waste in some cases), in comparison to materials like plastics, combustion is an ineffective means of disposal of or energy recovery from food waste [27]. The positive correlation between the proportion of food waste and moisture content of MSW in our datasets also confirms the unsuitability of incinerating moisture-rich food waste for energy recovery.…”

“…Predicting the LHV of MSW is one of its applications [3,22]. ANN is able to straightforwardly capture non-linear relationships between dependent and independent variables [3,22,27], as they avoid the need to identify an appropriate data-fitting function before the models can be constructed [24,28]. ANN models also allow for the inclusion of multiple inputs (or variables) and model adjustment of the models when new datasets are input [24].…”

“…Some researches indicated that most of the LHV of MSW was contributed by the paper, food (or organics in some cases) and plastics in the mixture, with this accounting for over 70% of the MSW on a dry basis [2,4,27]. The contribution of the remaining ingredients is statistically insignificant and may be neglected in the model.…”

“…The epoch (or number of iterations) varies from 200 to 4000, and the learning rate varies from 0.01 to 0.15 [3,13,22,27]. According to Dong et al, changes in the sum-squared error of the ANN are minimal beyond 800 epochs [22].…”