摘要
针对少数据、小样本序列的预测问题,在分析常规灰色GM(1,1)预测模型缺点的基础上提出了优化的算法模型,将优化后的预测方法推广为多参数预测.首先,建立了利用最新量测量进行初始化的预测模型,然后通过新陈代谢的方法利用新信息代替旧信息实现等维的模型预测,同时引入衰减记忆最小二乘法对新旧信息进行加权处理,背景值以归一化的平均相对误差作为精度检验标准,采用粒子群算法自适应寻优.最后,通过对某型惯性测量单元(IMU)的标定参数稳定性进行预测,预测结果平均相对误差降低了6%~58%,表明预测方法可以应用于IMU标定参数的长期稳定性预测.
The application of measured short-term data to the prediction of long-term stability of weapon system is significant to shorten the production cycle of weapons. Considering such prediction problems as inadequate data and small sample sequence, optimized algorithm model was presented based on the drawback analysis of GM(1,1) prediction model. The optimized prediction methods were generalized as multiparameter prediction. At first, the model which used the latest measured data for initialization was estab- lished, followed by replacing the old information with the latest through metabolic approaches to realize equal dimension model predication. In addition, fading memory recursive least squares method was adopted for weighted handling of old and new information. The normalized mean relative error was used as accuracy test standard for background value and particle swarm optimization algorithm was adopted. Finally, the calibration parameters stability of a certain type of inertial measurement unit (IMU) was predicted, and the average relative error of the prediction results was reduced by 6%-58%. The results indicate that the prediction method can be applied to the long-term stability of IMU calibration parameters.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
2017年第8期970-976,共7页
Journal of Shanghai Jiaotong University
基金
国家自然科学基金(61004124)
中央高校基本科研业务费专项资金(3102014KYJD007)资助
关键词
灰色理论
新陈代谢
自适应
预测
粒子群算法
grey theory
metabolism
adaptive
prediction
particle swarm optimization