目的玉米青贮是东北地区奶牛场年需求量较大的青贮饲料,快速、准确地评估其营养指标,可为精准饲养配方的制定提供有力数据支持。方法本研究选取120 份玉米青贮样品作为定标集,30 份作为验证集。在前期已建立的不同样品状态（湿基、烘干粉碎）发酵玉米青贮干物质（DM）、淀粉预测模型基础上,进一步维护数据库,同时建立不同样品状态下发酵玉米青贮粗蛋白（CP）、中性洗涤纤维（NDF）、酸性洗涤纤维（ADF）的近红外预测模型,用于东北地区奶牛场日常发酵玉米青贮的营养检测。结果（1）所建立的不同样品状态发酵玉米青贮CP、NDF、ADF预测模型均能实现对发酵玉米青贮的营养预测,其中CP的实际预测效果优于其他营养指标;（2）烘干粉碎后的发酵玉米青贮营养成分预测模型,其实际预测效果优于湿基发酵玉米青贮预测模型;湿基发酵玉米青贮NDF预测模型需进一步收集数据并优化。结论本研究建立的发酵玉米青贮CP、NDF、ADF近红外预测模型,可应用于东北地区奶牛场日常发酵玉米青贮的实际预测生产实践,且烘干粉碎样品的预测模型效果更优。

[Objective] Corn silage is a silage with high annual demand in dairy farms in Northeast China. Rapid and accurate evaluation of its nutritional indexes provides strong data support for the formulation of precise feeding formulas. [Method] In this study,120 corn silage samples were selected as the calibration set,and 30 as the validation set. Based on the previously established prediction models for dry matter （DM）and starch of fermented corn silage in different sample states （wet base, dried and crushed）,the database was further maintained. Meanwhile,near-infrared prediction models for crude protein （CP）,neutral detergent fiber （NDF）,and acid detergent fiber （ADF） of fermented corn silage in different sample states were established for the daily nutritional detection of fermented corn silage in dairy farms in Northeast China. [Result]（1）The established prediction models for CP,NDF,and ADF of fermented corn silage in different sample states could all predict the nutrition of fermented corn silage,and the actual prediction effect of CP was better than that of other nutritional indexes.（2）The prediction model for nutritional components of fermented corn silage after drying and crushing had better actual prediction effect than that of wet-based fermented corn silage.The NDF prediction model for wet-based fermented corn silage needed further data collection and optimization. [Conclusion] The near-infrared prediction models for CP,NDF,and ADF of fermented corn silage established in this study can be applied to the practical prediction of daily fermented corn silage in dairy farms in Northeast China,and the prediction model for dried and crushed samples is more effective.