Machine learning-based optimisation of microalgae biomass production by using wastewater

Abstract

Microalgae biomass a valuable resource for energy source and fertilisers. Among all available media sources, wastewater serves as cost-efficient media for microalgae biomass production. However, improvements are still needed to increase microalgae biomass productivity in wastewater, particularly optimising cultivation parameters, including light intensity, CO2 content, and initial inoculum level. Machine learning algorithms can easily optimise the cultivation parameters by analysing the microalgae-based wastewater treatment dataset without experimental runs. In the present work, a machine learning algorithm, Decision Tree, was used to analyse microalgae-based datasets and predict different optimised combinations of descriptor variables leading to high growth rate and microalgae biomass production. The decision tree model presented 18 combinations of descriptor variables leading to increased biomass production with a general accuracy of 81.25%. Seven of them were experimentally verified by using two newly isolated strains that predicted a 10% error after experimental verification. Models also revealed that the medium's initial inoculum level and nitrogen/phosphorus ratio highly influence microalgae growth as compared to other parameters. Results obtained in the analysis can be easily implemented at pilot and industrial scales without any laboratory experiments.