A non-intrusive load monitoring-enabled framework for load scheduling in the dairy industry

Tools

Vavouris, Apostolos and Stankovic, Lina and Stankovic, Vladimir (2025) A non-intrusive load monitoring-enabled framework for load scheduling in the dairy industry. Applied Energy, 398. 126456. ISSN 0306-2619 (https://doi.org/10.1016/j.apenergy.2025.126456)

[thumbnail of Vavouris-etal-AE-2025-A-non-intrusive-load-monitoring-enabled-framework-for-load-scheduling]
Preview
 Text. Filename: Vavouris-etal-AE-2025-A-non-intrusive-load-monitoring-enabled-framework-for-load-scheduling.pdf
Final Published Version
License: Creative Commons Attribution 4.0 logo
Download (5MB)| Preview

Abstract

The agricultural industry is an important contributor to CO2 emissions, with dairy accounting for over 15 % of total agricultural output in the United Kingdom (UK). In line with the United Nations’ Sustainable Development Goals for responsible consumption, and affordable and clean energy, decarbonisation of agriculture is being prioritised around the world, with integration of renewables, energy storage systems, and load flexibility widely recognised as viable solutions to achieve net-zero. Analysis and optimisation of energy-consuming agri-processes remains a huge challenge — compared to residential and commercial buildings — due to non-standardised equipment and the emergence of on-site renewables. In contrast to previous studies that are narrow-focused, do not involve end-users, do not consider the diverse and largely varying day-to-day energy-intensive activities, or are not applicable at scale, this article proposes a novel, deep learning-based, data-driven, modular non-intrusive load monitoring (NILM)-enabled approach, where context is set through co-creation with farms and agritech. The proposed approach enables accurate and scalable load disaggregation at very-low frequencies (30-min), through transfer learning, and scheduling of energy-consuming processes, which minimises, simultaneously, total electricity import and carbon footprint, based on renewable production prediction, and granular regional carbon footprint forecasting. Findings from three small to medium/large-scale dairy farms in the UK with renewables and diverse non-standardised dairy equipment demonstrated that through the completely non-intrusive and scalable co-created load scheduling approach based on identified flexible loads, utility is preserved under a very-low frequency (30-min) disaggregation scenario. The proposed system achieves electricity cost and carbon footprint reduction of over 30 % compared to current energy practices on the three farms, and paves the way for completely non-intrusive/no capital investment NILM-enabled systems for the agriculture industry.

ORCID iDs

Vavouris, Apostolos ORCID logoORCID: https://orcid.org/0000-0002-6138-7865, Stankovic, Lina ORCID logoORCID: https://orcid.org/0000-0002-8112-1976 and Stankovic, Vladimir ORCID logoORCID: https://orcid.org/0000-0002-1075-2420;
CORE (COnnecting REpositories)