从农场到餐桌的科学:以实现未来弹性生产的乳业数据挖掘  

Farm-to-table Science:Dairy Data Mining for Future Resilience

作　　者：王蕾(编译)[1] 郭凯军[1] WANG Lei;GUO Kaijun(Beijing University of Agriculture,Beijing 100096)

机构地区：[1]北京农学院,北京100096

出　　处：《中国乳业》2024年第11期50-56,共7页China Dairy

基　　金：北京市家畜创新团队项目(BAIC05-2024)。

摘　　要：自动挤奶系统(AMS)和精确畜牧业(PLF)系统提供的大量多维度数据,有助于探索现实世界的抗逆性状和可持续农业战略。AMS自20世纪90年代在荷兰问世以来,在北欧地区的应用日益广泛,由机器人收集的牛奶总量约占总产量的1/3。北欧地区主要品牌的牧场管理系统(FMS)仅能报告当前数据,未能涵盖对研究牛群遗传、行为和环境至关重要的历史信息。本研究介绍了瑞典农业科学大学(SLU)的奶牛数据基础设施——SLU Gigacow系统,该系统从某些瑞典农场收集数据。每个农场的FMS每晚向SLU Gigacow发送报告,该系统统一记录并将数据收集到中央数据库。收集的记录包括数千头奶牛的挤奶统计数据、健康事件、交易数据和SNP基因型,研究人员可通过结构化查询语言(SQL)或R语言查询并访问这些记录。SLU Gigacow还整合了来自瑞典全国奶牛登记处的数据,包括参与农场奶牛的系谱信息和牛群转移情况。SLU Gigacow的纵向观察(首批数据收集于2020年)将基因型、表型和动物福利联系起来,旨在加快乳业科学试点研究,并提供一个活跃于商业环境的奶牛大数据集。用Python 3(比弗顿,美国)数据收集编写软件可从多个版本的DeLaval DelPro(帝波罗公司推出的一款牧场管理软件系统)收集数据模块,并扩展到任何具有图形用户界面的FMS,且在大多数客户的操作系统运行。使用SSIS,即SQL Server集成服务协调解决语言、FMS版本的差异后,数据被存储在SLU Gigacow维护的数据库中,使用SSIS。通过与瑞典乌普萨拉的瓦克萨瑞典公司(Växa Sverige AB)达成协议,参与农民可使用由荷兰阿姆斯特丹生产的包含45K EUROG MD微珠芯片,对大量动物进行基因分型。目前,该数据库包含17000余头牛、300余万次挤奶、2969个SNP基因型的信息。交叉参考数据可用于各种目的,包括应激反应和抗逆性状研究。虽然SLU Gigacow旨在从瑞典农场收集数�Adoption of automated milking systems(AMS)and other precision livestock farming(PLF)systems offers access to large,multidimensional data that allow exploration of resilience traits and sustainable farming strategies in real-world scenarios.Since their inception in the Netherlands in the 1990s,AMS have seen increased adoption Adoption of automated milking systems(AMS)and other precision livestock farming(PLF)systems offers access to large,multidimensional data that allow exploration of resilience traits and sustainable farming strategies in real-world scenarios.Since their inception in the Netherlands in the 1990s,AMS have seen increased adoption in the Nordic countries,with around a third of the total milk production collected by robots.The major brands of farm management systems(FMS)in the Nordic region are only configured to report data as a current overview,discarding older information that is vital to studies of the herd’s genetics,behaviour,and environment.In this work,we present the infrastructure for dairy cattle data at the Swedish University of Agricultural Sciences,Gigacow(SLU Gigacow)that collects data from a set of Swedish dairy farms.Each farm’s FMS sends nightly reports to SLU Gigacow,where records are harmonised and collected in a central database.Collected records include milking statistics,health events,traffic data,and SNP genotypes for thousands of cows,and are made accessible to researchers through SQL or R queries.SLU Gigacow also integrates data from the Swedish national cow registry,including pedigrees and herd transfers for cows resident at participating farms.SLU Gigacow’s longitudinal observations(first data collected in 2020)link genotype to phenotype and animal welfare with the goal of accelerating pilot studies in dairy science,as well as providing a big dataset from cows in active,commercial settings.The data collection software written in Python 3(Beaverton,USA)has modules that enable collection from several versions of DeLaval DelPro(Tumba,Sweden),and can be extended to any F

关 键 词：乳品科学 大数据 抗逆性状 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论] F313[自动化与计算机技术—计算机科学与技术] F416.82[经济管理—产业经济]