Predictive accuracy and characterisation of bio-oil yield from pyrolysis of Cocos nucifera: A comparison of traditional RSM and hybrid models

Onokwai, Anthony O
Akuru, Udochukwu B.
Desai, Dawood A.

International Journal of Renewable Energy Development
Vol 14, No 6 (2025): November 2025

The pressing demand for renewable energy has made biomass a quintessential alternative to fossil fuels. This study aims to develop and compare predictive models for optimising bio-oil yield from the intermediate pyrolysis of Cocos nucifera, utilising response surface methodology with the central composite design and hybrid models (PSO-ANFIS and GA-ANFIS). It seeks to characterize the bio-oil yield to investigate its quality for use as a biofuel. An experimental run was performed by varying pyrolysis operating parameters, namely, temperature (300–700°C), heating rate (6–30°C/min), residence time (5–25 minutes), particle size (0.5–4.5 mm), and nitrogen flow rate (10–50 mL/min).  Hybrid models (PSO-ANFIS and AN-FIS-GA) were used to predict the bio-oil yield to identify the most robust model. An optimum bio-oil yield (52.17 wt.%) was attained at a temperature, heating rate, residence time, particle size, and nitrogen flow rate of 510.2°C, 10.5°C/min, 5.2 minutes, 0.3 mm, and 17.3 mL/min, respectively.  The study shows that its hybrid models are scalable and outperform traditional techniques (RSM) in terms of predictive accuracy and computational efficiency. The GC-MS analysis identified over 200 compounds in bio-oil, comprising mainly phenols, esters, and oleic acids, which confirmed its suitability for producing biofuels, lubricants, and pharmaceuticals. Also, FTIR analysis confirms functional groups of biodiesel, adhesives, and resins. The PSO-ANFIS and GA-ANFIS models accurately predict the bio-oil yield, with the PSO-ANFIS model outperforming the other models with an R² of 0.994 and RMSE of 0.449 during the test phase, representing a two- to three-fold improvement over traditional RSM. Unlike conventional empirical models, the hybrid approach improves predictive accuracy and reduces the number of required experiments and computational errors, enabling real-time adjustments to the pyrolysis process, thereby advancing pyrolysis research and bio-oil optimization. This research is highly relevant for improving waste-to-energy production in regions where Cocos nucifera residues remain abundant, especially in emerging economies.