AI researchers creating unique model for circular food production

Sunny summers, blistering winters – and residual heat from data centres. These are the ingredients that researchers from Luleå University of Technology will use to create an intelligent recipe for circular food production systems.

– This project allows us to lay foundations for strong collaboration with innovative land-based food production in Norrbotten. We are the first project aiming at a complete circular system with least negative effect on the environment, says Ali Usman, machine learning researcher at Luleå University of Technology.

The project, which will find new models for land-based food production in Norrbotten, has several components: artificial intelligence, the geographical location of Boden, circular use of nutrients in a closed food production system, and residual heat from, for example, data centres.

– We want to study the energy flows needed for food production, to better understand how residual heat from similar facilities can be used, referring to H2 Green Steel's establishment in Boden, says Michael Nilsson, project manager at Luleå University of Technology

A balanced diet

This is how everything is connected: In Boden, major investments with data centres and a steel plant has been made or is underway. These emit residual heat, a valuable source of energy, which could, for example, be used to grow vegetables or fish. At the present, there is no ready-made solution for this, partly because the energy flow shifts. For this to work, there need to be intelligent systems that can quickly adapt to these changing conditions.

– The project will reach a model of energy flows. The end goal is that the knowledge in the future can be applied practically in a full-scale facility, Michael Nilsson explains.

The project's theoretical ecosystem, which in the future could be located downstream of a data centre or hydrogen plant, consists of four biological components: microalgae, insects, vegetables, and fish. Microalgae and insects will compose nutrients for a sustainable fish feed. Existing pilot experiments and data provide information on how energy is transferred between the layers in different situations. Based on data, the AI ​​model can in turn, with the help of deep learning, create models of how and when the residual heat is to be used - and thus have a closed system, Ali Usman explains:

– The AI models are intended to simulate, within seconds, complicated physical processes which would take hours for computation, if the exact solution would be intended.

Deep learning

The hot summers and cold winters in northern Sweden can represent many different conditions, which are used in the survey. A model of how and when the residual heat is to be used will therefore be applicable to a Nordic climate.

– The research group develops machine learning methods that simulate the flow of energy and nutrition in a very complicated environment with insects, fish, microbes, and algae production with heat from data centres. By preserving nutrients in a circular system and combining with the energy from the data centre, we can make a difference: sustainable fish, and vegetables, made in Boden, Ali Usman concludes.