Lang
A team of Nigerian researchers has developed an advanced computational framework aimed at enhancing renewable energy production from oil palm waste, by integrating anaerobic digestion experiments with artificial intelligence and optimization technologies.
Published in the journal Biomass Conversion and Biorefinery, the study focused on converting oil palm empty fruit bunches (OPEFB), a lignocellulosic agricultural residue generated in large volumes during palm oil processing. The research highlights how this waste, often incinerated or discarded, can be transformed into a valuable clean energy resource while supporting a circular bioeconomy.
The project was led by Dr. Idowu Olugbenga Adewumi, a lecturer at the Federal College of Agriculture in Ibadan. The team investigated anaerobic digestion as a sustainable pathway for converting OPEFB into methane. The principal technical challenge lies in the complex structure of lignocellulosic biomass, where lignin impedes microbial access to cellulose and hemicellulose, limiting conversion efficiency.
To address this challenge, the researchers designed an integrated framework combining experimental measurements, machine learning, explainable artificial intelligence, and evolutionary optimization. The analysis covered key operational parameters: temperature, pH, organic loading rate, hydraulic retention time, and the carbon-to-nitrogen ratio.
The experimental programme generated 5,452 measured observations from laboratory-scale anaerobic digestion trials conducted between January and November 2025. Methane yields across the experiments ranged from 181.44 to 282.42 mL CH4 g⁻¹ VS, with an average yield of 231.89 ± 15.22 mL CH4 g⁻¹ VS.
Explainable AI analysis identified temperature as the most influential variable for methane output, followed by cellulose content. The researchers noted that elevated temperatures enhance microbial activity, accelerate substrate breakdown, and support methanogenic microorganisms.
Five predictive models were benchmarked in the study: Multiple Linear Regression (MLR), Artificial Neural Network (ANN), Random Forest (RF), Support Vector Regression (SVR), and Gradient Boosting Regression (GBR). Multiple Linear Regression delivered the strongest predictive performance, with an R² of 0.5513, RMSE of 10.14 mL CH4 g⁻¹ VS, MAE of 8.17 mL CH4 g⁻¹ VS, and MAPE of 3.54%.
The team then applied evolutionary optimization algorithms, specifically the Genetic Algorithm and Particle Swarm Optimisation, to identify the conditions that would maximize methane output. The model predicted optimum conditions at a temperature of 54.87 °C, pH of 7.18, organic loading rate of 3.62 g VS L⁻¹ day⁻¹, hydraulic retention time of 28.41 days, and carbon-to-nitrogen ratio of 27.83. Under these parameters, projected maximum methane production reached 281.84 mL CH4 g⁻¹ VS.
The researchers emphasized that the integrated framework delivers a transparent decision-support system for improving anaerobic digestion efficiency and accelerating the deployment of sustainable waste-to-energy technologies. The findings carry particular weight for Nigeria, where agricultural residues from palm oil production represent a significant untapped resource for renewable energy generation and environmental protection.
By merging biotechnology with artificial intelligence, the study illustrates how digital technologies can unlock more efficient pathways for converting agricultural waste into valuable energy resources.
Source: Tribune Online