奶牛泌乳曲线数学模型拟合和早期预报

RESEARCH ON FITTING MATHEMATICAL MODELS OF LACTATION CURVE IN DAIRY CATTLE AND THEIR EARLY STAGE PREDICTION

本研究根据北京地区黑白花奶牛500个完整泌乳期的产奶记录资料研究了Wood模型、改进多项式模型、回归模型的拟合以及回归模型的预报情况。拟合结果表明,利用Wood模型来估计北京地区黑白花奶牛产奶量是可行的,但在取样间隔天数较大时估计奶量与实际奶量间有较大的偏差。鉴于此,作者提出了改进多项式模型,采用了ALI的回归模型,按对Wood模型进行拟合的同样方法对这两个模型进行了拟合,拟合精确度都较Wood模型有所改进。取样间隔天数大时,改进多项式模型的拟合精确度提高的幅度较大。回归模型的预报奶量与实际奶量呈较一致的变化规律,这对早期选种具有一定的实践意义。

Daily milk records from 500 Holstein lactations in Beijing area were regressed on time using 3 mathematical models. The models included Wood model: yt=anbe-ct, modified polynomial model:yt = a +bt + ct2 + dlnt +etsin ( 2πt/1 ) and regression model:yt = a + b ( t/1 ) + c ( t/1 )2 + dln ( 1/t ) +e[ln(1/t)]2. In these models, yt is daily milk yield at time t, a, b, c, d, e are constants determined by least squres, and 1 is the number of comlete lactation. The results of fitting showed Wood model might be practical in Beijing area. Sampling milk yield records each 30-day and each 60-day, the modified polynomial model gave the best fit to the data. Sampling each 10-day, good estimations were observed for all these models. The regression model was used to predict complete lactation from part-lactation yield. The absolute error,relate error, and R2 value of prediction were 456.7kg, 5.80%, and 0.984270 respectively. The predict value obtained from the regression model has a strong high correlation, r= 0.936146, with milk yield of actual complete lactation.