< Back to previous page

Project

Precision Dairy Nutrition (MELKDATA)

Main research question/goal
This project links dairy cow management (in casu the rationing of lactating cows) to tools from the computer sciences. Specifically we are creating a prediction model based on supervised learning algorithms. The idea is to express the milk production of each cow in the herd as a function of the kinds and quantities of feed intake by the different animals. Once the model is fully functional it will become possible for the dairy farmer to feed the animals precisely and individually without spending more time per animal. The dairy farms in Flanders and in Europe are in a transitioning phase, which is leading to increased relevance of such predictive computer models. Maintaining current profit margins without being hindered by manual labor has created the need for maximal efficiency through automated systems.

Research approach
The ILVO cows (120 -140 lactating cows housed in the ILVO research barn) are our research material. They are individually and intensively monitored (feed intake, activity, milk production, milk composition,…). We store the registered data in a relational database. This extensive dataset, full of time series, is used to detect and document the relationships between the observations. As such it becomes possible to make predictions with regards to later observations. We use the techniques of supervised learning algorithms. When the relationship is demonstrated and modeled, we test these hypotheses by using new observations in the research stable, such that the milk production can be optimized further. Not every sensor of observations will prove to be equally important in such models. We select the sensors that give us the most relevant data and use those to advise the commercial farms. At our partner dairy farms we place the relevant sensors and actuators. This enables us to validate the proposed model.

Relevance/Valorisation
Advice to the dairy farmer about the optimal portion concentrate, roughage, etc. for an individual cow will become much more precise on the basis of the algorithms and the validated model. We expect that, on the basis of this methodology, similar models will be developed to advise the dairy farmer regarding their cow management (e.g., disease management). Especially the larger companies will prosper with such a form of automation and optimization of the ration. The dairy sector is experiencing a scale-up, so the group of potential users is on the rise. The scale-up is mainly attributed to the predicted worldwide increase in dairy products (OESO) and the end of the European milk quotas (April 2015).
Date:26 Jan 2015 →  25 Jan 2019