A research initiative to develop low-cost IoT solutions for monitoring bulk milk cooling tanks in real-world dairy farms.
Can you really know that the entire profitability of a dairy farm
goes through a refrigerated steel tank — with no monitoring?
Most farms have zero real-time visibility into what happens inside their bulk milk tank. This research addresses that gap with affordable, deployable IoT hardware validated in real farm environments.
The LactoKeeper research platform comprises two complementary prototypes, both validated in real dairy farm deployments across multiple farms in Castilla-La Mancha.
A self-contained floating sensor that records temperature and motion data to a MicroSD card. No network required — designed for on-demand veterinary audits and remote farm deployments.
The connected research platform. A LoRaWAN gateway aggregates data from multiple tanks in real time, enabling cloud-based monitoring, energy consumption analysis, and AI-powered stage detection.
The LactoCloud dashboard was developed as part of the research to validate monitoring hypotheses and visualise field data from real farm deployments. Screens shown are from live research deployments.
A core research contribution is a machine learning model that automatically identifies every operational stage from raw sensor data — validated against manually labelled field recordings.
The model processes gyroscope magnitude, temperature curves, and temporal patterns to classify each moment in the tank's lifecycle. Any anomaly — a missed wash, irregular cooling ramp, unexpected agitation — is detected and flagged automatically.
Energy consumption data is analysed in parallel, correlating electrical patterns with operational phases. This allows the research to identify inefficiencies, detect abnormal consumption during cooling or agitation, and build a baseline for energy optimisation in dairy storage.
Air quality readings from the tank room and electrical consumption data each come from dedicated sensor nodes — separate from the floating device. Together they provide environmental and energy context that complements the in-tank data, enabling a more complete picture of storage conditions and equipment efficiency.
Intelligent deep-sleep scheduling on ESP32 enables months of autonomous deployment — a critical requirement for real farm validation without infrastructure changes.
The floating device integrates 6-DoF IMU and temperature sensing. Separate dedicated nodes handle air quality in the tank room and electrical energy consumption — each optimised for its specific measurement context.
Low-power wide-area network covering entire farms without Wi-Fi infrastructure. One gateway handles multiple tanks simultaneously across different buildings.
Real-time measurement of tank electrical consumption — enabling research into correlations between energy usage, operational phases, and equipment efficiency.
Environmental sensors in the tank room capture temperature, humidity, and air quality parameters — providing context for tank behaviour and storage conditions.
Every event is timestamped and stored. Field data from multiple farms provides the empirical basis for the research findings and AI model training.
LactoKeeper has been validated across multiple dairy farms in Castilla-La Mancha. The research findings have been submitted for publication and are available as a preprint on SSRN.
A multidisciplinary research team from the Universidad de Castilla-La Mancha — combining expertise in embedded systems, IoT, data engineering, and smart agriculture.
Developed in collaboration with clinical veterinarians specialised in milk quality. The research has been validated with veterinary professionals with decades of field experience in dairy herd health and storage protocol compliance — ensuring the system addresses the real informational needs of quality auditors and farm advisors.
Whether you're a dairy farmer, a veterinarian, a researcher, or an institution interested in collaborating — we welcome enquiries about field validation partnerships, research collaboration, or access to the dataset.