Improving scale-up efficiency and deployment speed for biorefineries

Biorefineries allow production of multiple products from different feedstocks, with outputs such as high-value pharmaceutical or nutraceutical compounds to lower commodity products from energy vectors (solid, liquid or gaseous fuels) to biochar, with potential added value to treat water or utilise waste streams, as well as offering carbon capture and utilisation solutions.

The input to biorefineries relies on available feedstocks, such as energy crops, agricultural waste, refuse derived fuel or even microalgae. A primary question for their development is the feedstock sustainability and whether different feedstocks are available through the year to support commericalisation at scale. Feedstock variety can also lead to process and scaling problems resulting in complex pathways towards technology process selection (eg, thermochemical, fermentation or biotechnology) and valuable products which will generally vary with feedstock. Demonstration of biorefinery steps and processes at laboratory scale provides important information for process viability and scaling but finding deployable strategies, processes and systems that can be scaled commercially and rapidly whilst allowing net zero or negative carbon emissions remains a challenge.

The work will focus on developing models based on machine learning and artificial intelligence to identify viable pathways for biorefinery development, scaling and deployment, including adaptive approaches to back-fill incomplete data sets. Strategies will be developed to identify optimal machine learning approaches to these multifactorial problems from laboratory to pilot and commercial biorefinery plants, identifying feedstock options from growth to product along with technical process decisions, deployment of biorefinery concepts and how machine learning can be combined along with systems control, downstream processing and optimisation.

This project is led by Ian Watson.