top of page

AI for the Decarbonisation of the Dairy Sector: Heat Recovery and Energy Harvesting

Prof Lina Stankovic, University of Strathclyde; Dr Sean Cullen, Queen's University Belfast; Richard Hey, RHD Scotland Ltd

LStankovic AI for decarbonisation of dairy sector.jpg

​

Interview date

​

TBC


Apply for this studentship
​
See our Application Page.

​

​

Research Aims

​

As highlighted in the DEFRA Agri-climate report 2023, the UK agricultural sector has seen a steady decrease in NOx and methane emissions between 1990 and 2021, but significant 22% increase in CO2 emissions during the same period. Furthermore, according to the DEFRA report, there are still relatively large uncertainties in estimating agricultural emissions and quantifying actual CO2 production from various milking machinery and electricity-consuming processes due to the lack of integration of granular spatio-temporal energy production and generation mixture data. There is also increased financial pressure on the dairy industry, mainly due to soaring energy prices and fast adoption of automation and electrification of processes, moving away from fossil fuels, such as automated milking robots. Sustainable approaches to harness renewable energy sources, such as wind, solar and hydro, as well as waste heat recovery systems are being embedded into farms. Novel solutions such as excess renewable energy production being stored in the form of ice with the use of compressors or warm milk cooling plates and mattresses used for heat exchange and recovery are being introduced in different farms. However key findings from our recent projects with farms via DDC IVs has demonstrated that without an AI-driven energy management system that incorporates consumption, generation and heat recovery for storage and flexibility services, fossil fuel use and therefore CO2 emissions reduction is not possible. The project will explore: 1) how AI data-driven modelling can improve physical thermodynamic models by learning key trends within the collected physical quantities and uncover anomalous activation patterns; 2) how regression-based ML modelling for day- and week-ahead forecasting of energy generation from RES and heat recovery systems can be improved; and, 3) How deep reinforcement learning based recommender systems for energy management incorporating the above developed models can be developed. The PhD student, ideally with an electronics, computing or physics background will be trained in a range of supervised and unsupervised machine learning models for anomaly detection, forecasting and recommendation systems as well as hands-on experience with historical and ongoing energy consumption and generation data from a range of dairy farms. The student will also benefit from qualitative approaches of research by design and circularity in the construction and resource management of the farms.

​

bottom of page