A novel machine learning-based approach for prediction of nitrogen content in hydrochar from hydrothermal carbonization of sewage sludge

Abstract In this work, 138 datapoints, including elemental composition and ultimate analysis of various types of sewage sludge, and the hydrothermal carbonization reaction conditions, are used to develop a prediction model for the nitrogen content of the hydrochar. The results suggested that a two-layer feedforward neural network with five (05) neurons in the hidden layer can accurately predict the nitrogen content of the hydrochar based on the reaction temperature and the contents of nitrogen, carbon, volatiles and fixed carbon in the feedstock. Over 100 runs, the R2 and RMSE are in [87.547–99.097%] and [0.243–1.431] wt.% (db), respectively. Moreover, a statistical and regression analysis revealed that the sewage sludge-N is the main contributor to the hydrochar-N. Mostly, 40–70% of sewage sludge-N goes to hydrochar-N. The results are consistent with previous experimental reports, and this model can be used to predict the sewage sludge-derived hydrochar-N.