(523f) Critical Material Attributes of Algae Biomass for Successful Preservation in Storage Identified through Machine Learning Models

The biochemical composition of algae biomass impacts downstream processing of algae biomass to fuels. Algae biomass with a greater proportion of carbohydrates and lipids are desirable for maximizing fuel yield and quality in both biological and thermochemical conversion approaches. Yet maximizing productivity of algae in cultivation yields biomass with a greater proportion of protein than either carbohydrate or lipid, limiting yields in biochemical conversion approaches and requiring more extensive upgrading of bio-oil obtained from thermochemical processing. In addition to affecting conversion, algae composition could impact other important aspects of processing post-harvest algae biomass. Wet anaerobic storage is essential to ensuring continuous and timely supply of high-quality algae feedstocks to a biorefinery and the stability of algae biomass in storage is likely affected by its composition. To better understand the relationship between composition and storage stability we have utilized machine learning and 450 storage experiments to correlate material attributes of algae biomass at harvest with storage outcomes and to identify those that are critical to successful storage. Prediction accuracy of various machine learning models including linear regression, artificial neural network, KNN, and Random Forest in predicting dry matter loss and biochemical compositions were evaluated using coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE). Utilizing model-predicted critical material attributes, low-cost preservation approaches were devised to ensure successful storage regardless of biomass composition at the time of harvest. The effect of model-derived treatments on dry matter loss and the algae biomass microbiome are discussed.