Smart modeling of oil shale pyrolysis: Impact of feed composition and thermal parameters

Abstract

The development of oil shale resources is gaining global importance due to increasing energy demand. Pyrolysis is the most economical and feasible method for extracting oil from shale, yet its performance is difficult to predict due to complex reaction mechanisms, uncertainties in Arrhenius-type kinetics, and the often-overlooked role of heating rate. This study presents a novel application of artificial intelligence (AI) to predict oil shale pyrolysis behavior. Using elemental composition (C, H, N, O, S wt%), temperature, and heating rate as inputs, AI models were trained to estimate residual solid content after pyrolysis. A dataset of 457 thermogravimetric analysis (TGA) experiments was used, with an 80:20 train-test split. Three AI models—Artificial Neural Network (ANN), Adaptive Neuro-Fuzzy Inference System (ANFIS), and Least Squares Support Vector Machine (LSSVM)—were developed for comparison. ANFIS outperformed the others, achieving R² = 0.9706, RMSE = 5.6, and SD = 0.030. Graphical evaluations (residuals, scatter plots, Box-Cox, and cumulative error plots) supported its robustness. Sensitivity analysis based on relevancy factors showed that increased heating rate reduces pyrolysis efficiency, while higher temperature and elemental contents positively influence performance. Temperature was the most significant factor (r = –0.6766), indicating its central role in process optimization. This work demonstrates the effectiveness of AI in modeling oil shale pyrolysis and offers a valuable tool for improving operational strategies.