Periodic Reporting for period 2 - HAYTECH AI (Hay data analytics & fermentation control to increase dairy farm milk yield)
Periodo di rendicontazione: 2024-02-01 al 2025-10-31
The HAYTECH AI project (Grant 190159907) addresses a major challenge in the dairy and hay-production industry: nutrient losses and fire risks caused by uncontrolled fermentation in hay bales. Up to 30 % of stored hay can spoil, leading to reduced milk yield, increased waste, and higher CO2 emissions.
Quanturi Oy develops HAYTECH AI, a system combining wireless IoT probes with AI-driven analytics to predict hay quality and identify risks before spoilage occurs. Each probe measures internal bale temperature and transmits the data to a secure cloud platform (HAYTECH.app). Machine-learning models integrate these readings with agronomic and environmental data to deliver accurate predictions of nutritional quality: Crude Protein (CP), Neutral Detergent Fibre (NDF), and Relative Feed Value (RFV).
By enabling farmers to monitor hay condition in real time and plan optimal feeding strategies, HAYTECH AI helps increase milk yield by 1.5 – 2 %, reduces hay losses by ≈ 10 %, and lowers the carbon footprint per litre of milk.
Originally conceived to include a fermentation-control reagent mechanism, the system evolved after field trials and reviewer recommendations into a predictive, preventive approach focused on data analytics rather than chemical intervention. This pivot improved scalability, safety, and market acceptance while preserving the project’s innovative edge.
Quanturi Oy develops HAYTECH AI, a system combining wireless IoT probes with AI-driven analytics to predict hay quality and identify risks before spoilage occurs. Each probe measures internal bale temperature and transmits the data to a secure cloud platform (HAYTECH.app). Machine-learning models integrate these readings with agronomic and environmental data to deliver accurate predictions of nutritional quality: Crude Protein (CP), Neutral Detergent Fibre (NDF), and Relative Feed Value (RFV).
By enabling farmers to monitor hay condition in real time and plan optimal feeding strategies, HAYTECH AI helps increase milk yield by 1.5 – 2 %, reduces hay losses by ≈ 10 %, and lowers the carbon footprint per litre of milk.
Originally conceived to include a fermentation-control reagent mechanism, the system evolved after field trials and reviewer recommendations into a predictive, preventive approach focused on data analytics rather than chemical intervention. This pivot improved scalability, safety, and market acceptance while preserving the project’s innovative edge.
By the end of the project, Quanturi finalised all technical, validation, and market activities:
• Hardware & Platform: The probe and base-station architecture were redesigned for scalability and validated. Latency < 60 s, wireless range ≈ 1 km, battery life > 18 months, and > 2 400 probes per base station were demonstrated. Manufacturing cost fell 53 %.
• AI System: Predictive models achieved ≈ 85 % accuracy for CP and NDF values, and the Feed Value Indicator (FVI) was validated as a decision-support tool for ration optimisation.
• Pilots & Market Validation: Ten pilot farms in France, Germany, Italy, and the UK plus field sites in Australia validated usability and agronomic accuracy. Distribution partnerships were established with John Deere (U.S.) and Farmscan (Australia).
• Exploitation & Dissemination: HAYTECH AI was showcased at Agritechnica 2023, World Ag Expo 2024, and numerous webinars and fairs. Five demo days (> 250 participants) and online outreach (+ 45 % audience growth) expanded awareness.
All work packages achieved 100 % completion, and the system reached TRL 8 – System Complete and Qualified.
• Hardware & Platform: The probe and base-station architecture were redesigned for scalability and validated. Latency < 60 s, wireless range ≈ 1 km, battery life > 18 months, and > 2 400 probes per base station were demonstrated. Manufacturing cost fell 53 %.
• AI System: Predictive models achieved ≈ 85 % accuracy for CP and NDF values, and the Feed Value Indicator (FVI) was validated as a decision-support tool for ration optimisation.
• Pilots & Market Validation: Ten pilot farms in France, Germany, Italy, and the UK plus field sites in Australia validated usability and agronomic accuracy. Distribution partnerships were established with John Deere (U.S.) and Farmscan (Australia).
• Exploitation & Dissemination: HAYTECH AI was showcased at Agritechnica 2023, World Ag Expo 2024, and numerous webinars and fairs. Five demo days (> 250 participants) and online outreach (+ 45 % audience growth) expanded awareness.
All work packages achieved 100 % completion, and the system reached TRL 8 – System Complete and Qualified.
HAYTECH AI moves the hay-quality sector from reactive observation to predictive management.
Key results include:
• A validated AI algorithm predicting hay nutritional composition from agronomic data (RMSE ≈ 3 %).
• The Feed Value Indicator, transforming complex forage parameters into a simple score usable by farmers and feed advisors.
• An industrial-grade IoT platform linking hardware, cloud analytics, and user interface within one ecosystem.
• A 50 % reduction in sensor cost, enabling wide adoption in smaller farms.
• A unique, data-driven prevention model reducing hay waste and increasing milk-yield efficiency.
These results push the boundaries of digital agriculture by combining remote-sensing, AI, and livestock nutrition analytics in a single, validated system.
Key results include:
• A validated AI algorithm predicting hay nutritional composition from agronomic data (RMSE ≈ 3 %).
• The Feed Value Indicator, transforming complex forage parameters into a simple score usable by farmers and feed advisors.
• An industrial-grade IoT platform linking hardware, cloud analytics, and user interface within one ecosystem.
• A 50 % reduction in sensor cost, enabling wide adoption in smaller farms.
• A unique, data-driven prevention model reducing hay waste and increasing milk-yield efficiency.
These results push the boundaries of digital agriculture by combining remote-sensing, AI, and livestock nutrition analytics in a single, validated system.