Recruitment of long short-term memory for envisaging the higher heating value of valorized lignocellulosic solid biofuel: A new approach

Abstract

ABSTRACT The valorization of lignocellulosic wastes via the concept of bio-based circular economy to achieve the sustainable development goals of clean energy, safe life on land, and climate change mitigation is a worldwide scope nowadays. Lignocellulosic wastes are considered sustainable energy resources; consequently, it is crucial to find a cost-effective and time-saving method for predicting its higher heating value (HHV) to qualify its feasibility as a solid biofuel. In this study, the long short-term memory (LSTM) algorithm as a deep-learning (DL) approach has been applied in a pioneering step to calculate the HHV from 623 proximate analyses of various lignocellulosic biomasses. The relatively high value of the correlation coefficent of determination (R 2 0.8567) and low values of mean square error (MSE 0.67), root-mean-square error (RMSE 0.819), mean absolute error (MAE 0.597), and average absolute error (AAE 0.0319) confirmed the exceptional accuracy of the suggested LSTM model. Thus, recommending applying DL-LSTM as a new approach for building models since it provides an accurate prediction of HHV without the need for time-consuming and complicated experimental measurements or the conventional regression analysis and statistical modeling.