Artificial Neural Network Approach to Predict Biodiesel Production using Algae-Based Heterogeneous Catalyst

¹ Department of Mechanical and Biosystem Engineering, IPB University, IPB Darmaga Campus, Bogor, West Java 16680, Indonesia.
² Surfactant and Bioenergy Research Center (SBRC), IPB University, Indonesia
³ Department of Mechanical Engineering, Universitas Lambung Mangkurat, Indonesia
⁴ Department of Mechanical Engineering, Universitas Riau, Indonesia.
⁵ Research Center for Biomass and Bioproducts, National Research and Innovation Agency (BRIN), Indonesia.
⁶ Department of Forest Products, Faculty of Forestry, IPB University, Bogor 16680, Indonesia.
⁷ Universitas Pendidikan Indonesia, Bandung, Indonesia
⁸ Department of Civil and Environmental Engineering, A'Sharqiyah University, 400 Ibra, Oman.
⁹ GUST Engineering and Applied Innovation Research Center (GEAR), Gulf University for Science and Technology (GUST), Hawally, Kuwait.
¹⁰ School of Agricultural Engineering and Food Science, Shandong University of Technology, China.

^* Corresponding author: This email address is being protected from spambots. You need JavaScript enabled to view it.

Abstract

The rising dependence on fossil fuels has intensified environmental issues such as greenhouse gas emissions and resource depletion. Biodiesel offers a renewable alternative with lower emissions. However, conventional biodiesel production are sensitive to free fatty acids and water, causing soap. Heterogeneous catalysts derived from biomass provide a cleaner and reusable alternative. In this study, Ulva lactuca, a green macroalga with rapid growth and no need for arable land or fertilizers, was used as a sustainable source for catalyst preparation. This research integrates an U. lactuca-based heterogeneous catalyst with an Artificial Neural Network (ANN) to predict biodiesel yield under different process conditions. The objective was to develop a robust predictive model for biodiesel production from waste cooking oil. Transesterification was performed at 50–70 °C, catalyst loadings of 2–5 wt%, and reaction times of 60–180 min, with a fixed methanol-to-oil ratio of 6:1. The ANN, trained using the Levenberg–Marquardt algorithm in MATLAB R2022a, achieved an optimal architecture of 4–18–1. The model showed excellent predictive accuracy, with R values of 0.9989, 0.9969, 0.9980, and 0.9987 for training, validation, testing, and overall datasets, and minimum MSE values of 2.81 × 10⁻⁴. The highest experimental biodiesel yield of 0.96 mol mol⁻¹ closely matched the ANN-predicted yield of 0.97 mol mol⁻¹ at 60 °C, 90 min, and 4 wt% catalyst loading. These results confirm the ANN’s strong predictive capability and demonstrate its potential for optimizing biodiesel production using sustainable algae-based catalysts.