(375a) Physics-Informed Estimation of Thermodynamic Parameters of Biodiesel Production from Enteromorpha Compressa Microalgae

Due to the gradual depletion of the world's fossil fuel reserves, rising energy demand, and the detrimental environmental impacts of existing energy sources, there is an urgent need to identify sustainable alternatives [1]. Biodiesel has considerable potential to contribute to meeting this need, as it is biodegradable, and reduces total non-combusted hydrocarbon emissions by more than 90% [2]. Microalgae stand out as particularly promising feedstock for biodiesel production due to their high lipid content and fast growth rates [3]. The lipids extracted from microalgae go through the transesterification reaction to produce fatty acid methyl esters (i.e., biodiesel). However, because of the differences in the oil composition of different microalgae, the kinetic and thermodynamic parameters of the transesterification reaction associated with any type of microalgae differ [4]. These parameters are key in the design of biodiesel production operations; therefore, their precise evaluation is critical for improving the efficiency and effectiveness of the biodiesel production process. These kinetic parameters are typically estimated empirically from experimental data using black-box regression techniques. However, such empirical estimation often lacks precision and the integration of underlying physics, ultimately limiting the depth of understanding and interpretability of the derived parameters.

In this work, the physics-informed neural networks (PINNs) have been implemented to determine the kinetic and thermodynamic parameters associated with biodiesel production from Enteromorpha Compressa microalgae. The experimental data [6] from the transesterification of this microalgae cultivated in an isothermal batch reactor was used to construct the PINN model. The ordinary differential equations of the mass balances in the reactor are posed as a constraint on the neural network training to make the model physics-informed. This facilitated a precise behavior prediction to find kinetic parameters because physical laws governing the system constrain the space of acceptable solutions. The results show that the biodiesel production from this microalga is exothermic and non-spontaneous in nature. Furthermore, the kinetic and thermodynamic parameters were independently sought through the application of Evolutionary Algorithms, which demonstrated a consistent trend with the neural network implementation. Additionally, the optimum operational conditions were identified to achieve stable production with the highest possible biodiesel yield.

References:

[1] Ganesan, Devanesan, Aravindan Rajendran, and Viruthagiri Thangavelu. "An overview on the recent advances in the transesterification of vegetable oils for biodiesel production using chemical and biocatalysts."Reviews in Environmental Science and Bio/Technology 8 (2009): 367-394.

[2] Knothe, Gerhard, Christopher A. Sharp, and Thomas W. Ryan. "Exhaust emissions of biodiesel, petrodiesel, neat methyl esters, and alkanes in a new technology engine." Energy & fuels 20, no. 1 (2006): 403-408.

[3] Moser, Bryan R. "Biodiesel production, properties, and feedstocks." In Vitro Cellular & Developmental Biology-Plant45 (2009): 229-266.

[4] Singh, S. P., and Dipti Singh. "Biodiesel production through the use of different sources and characterization of oils and their esters as the substitute of diesel: a review." Renewable and sustainable energy reviews 14, no. 1 (2010): 200-216.

[5] Raissi, Maziar, Paris Perdikaris, and George E. Karniadakis. "Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations."Journal of Computational physics 378 (2019): 686-707.

[6] Suganya, Tamilarasan, Nagarajan Nagendra Gandhi, and Sahadevan Renganathan. "Production of algal biodiesel from marine macroalgae Enteromorpha compressa by two step process: optimization and kinetic study." Bioresource technology 128 (2013): 392-400.