针对很多预测案例中历史数据序列非等时距的特点,构建了非等时距的GM(1,1)预测模型,将序列的时间间隔作为乘子嵌入模型中,同时通过动态采用最新分量作为初始值、动态优化背景值和累积残差修正等方法,解决了非等时距预测模型长期预测精度不易控制的难题.将该模型应用于某发动机油液监测数据预测中,预测效果较好. Aiming at the characteristics of non-isochronous distance of historical data series in many prediction cases, a non-isochronous GM (1,1) prediction model is constructed, and the time series of the sequence is embedded into the model as a multiplier. At the same time, Initial value, dynamic optimization background value and cumulative residual error correction method to solve the long-term prediction accuracy of non-equidistance prediction model is not easy to control the problem.This model is applied to a certain engine oil monitoring data prediction, the prediction effect is better.