Investigation of pyrolysis potential of date seeds for bioenergy production: kinetic study by thermal analysis and activation energy prediction using advance predictive models

• Assessment of date seeds pyrolysis behaviour was complemented. • Model-free kinetics was employed to analyze the process kinetics. • Combined kinetics displays the Ea of 272.4 ± 11.8 kJ mol -1 having R 2 of 0.9948 • ANN, BRT, C&RT and MARS accurately predict the Ea for date seeds pyrolysis The pyrolysis potential of date seeds (DS) (an abundant agricultural residue that can support sustainable and resilient energy systems) as a renewable bioenergy feedstock was examined. Thermogravimetric analysis (TGA) of date seeds was performed from ambient temperature to 1000 °C under nitrogen at heating rates of 6, 9, 12, and 15 °C/min. The feedstock showed high volatile matter and a higher heating value (HHV) of 20.185 MJ/kg, confirming its suitability for thermochemical conversion. Isoconversional model-free methods such as Friedman (FR), Kissinger Akahira-Sunose (KAS), Ozawa-Flynn-Wall (OFW) and an advanced Vyazovkin (VZ) approach were applied over a conversion range of 0.2 to 0.8. A linear combined kinetics analysis gave an apparent activation energy (E a ) of 272.4 ± 11.8 kJ/mol with a correlation coefficient of 0.9948 with the reaction order of 7.75 ± 0.27. Four machine learning (ML) models, including Artificial Neural Networks (ANN), Classification and Regression Trees (C&RT), Boosted Regression Trees (BRT), and Multivariate Adaptive Regression Splines (MARS), were used to predict E a obtained from thermogravimetric data. The ANN achieved the best performance metrics, with a coefficient of determination (R2) of 0.985 and a Root Mean Squared Error (RMSE) of 3.84. The integrated kinetic and machine-learning framework provides reliable estimates of E a for DS pyrolysis. The predicted E a determines the temperature sensitivity of pyrolysis, setting the required heating rate, residence time, and temperature profile in the reactor. These results provide process-level input for reactor design, scaling-up, and optimizing bioenergy production from DS waste.