Data-driven dynamical modeling of dairy production oriented to herd management

• Published in: IFAC-PapersOnLine 2022, Vol.55 (10), p.1441-1446
• Main Authors: Rodriguez-Obando, Diego J., Castro-Rincón, Edwin, Castaño-Marin, Angela M., Montes-Vergara, José C., Rosero, Nadia
• Format: Article
• Language: English
• Subjects: decision-making; DIM; lactation curve; Particle Swarm Optimization; prognostics
• ISSN: 2405-8963; 2405-8963
• DOI: 10.1016/j.ifacol.2022.09.593

This paper presents a data-driven modeling of dairy production oriented to decision-making for herd management. Database was obtained between 2016 and 2021, and contains 62596 observations corresponding to information from up to 4 lactations for 107 animals from the herd at Obonuco Research Center (CI Obonuco) of AGROSAVIA. The CI Obonuco is located at the Colombian high tropic. The lactation curve is a mathematical tool for graphical representation over time of the biological process of milk yield, which is useful for predicting total production over a period of time. For the construction of lactation curves parametric and explicitly time-dependent models are generally used to fit data. In this paper we propose a parametric dynamical linear model whose output variable depends on its previous values and whose parameters are found by means of an optimization process. The Particle Swarm Optimization technique is used to solve the optimization problem. The model is validated from data. In addition, a general architecture is proposed to use historical data to increase the learning about dairy production system and improve its prognosis using few new data for decision-making.