AI-based monitoring of dairy cows’ feeding to reduce GHG emissions on dairy farms

Purpose and goal:

This project aims to leverage AI to create a sustainable and efficient approach to dairy farming. An IoT-monitoring system, with AI models deployed on the cloud, was installed in Vadsbo. Data from videos, management systems, and open sources was collected, providing a holistic view of feeding activities.

Two AI models were developed: one to measure feed intake and another to monitor cattle behavior. We identified the potential to build a predictive model for estimating methane emissions. Challenges included limited data granularity and occasional IoT system reliability issues. Expected results and effects:

The project successfully developed AI algorithms to measure feed efficiency regarding to GHG emissions in dairy farming and tested different IoT integration methods on a cattle farm. The IoT system was installed and effectively collected data. The AI models provided insights into cattle feeding behavior and paved the way for future advancements.

Key findings pointed out influncing factors like feed plan, age, and breed. Optimization determined conditions for maximizing milk production while minimizing methane emissions, adjusting feeding recipes and frequency. Approach and implementation:

The design and implementation process involved planing and installing IoT monitoring system, collecting data on cattle information, feed, milk production, and environmental conditions. After cleaning and preprocessing the data, exploratory data analysis was conducted to understand data distribution and relation ships. Feature engineering was performed to create predictive features for methane emissions.

AI models were chosen and trained using a split of the data into training and test sets. Model evaluation was done to ensured model robustness.