摘要
为了实现对测量仪器校准间隔的优化,对其校准数据进行建模,用GM(1,1)灰色预测模型预测参数的总体发展趋势,在此基础上利用BP神经网络对残差序列进行建模,通过训练补偿预测参数在总体趋势下的随机波动,从而得到校准数据的预测值。给出基于残差补偿的GM预测模型,对校准间隔进行动态优化,并通过实验对预测模型进行了验证。结果表明,此模型得到了较好的预测结果,既能预测总体趋势也能适应随机波动,并且简单易行,具有较强的普适性。
In order to realize the optimization of measuring instrument calibration interval , set up the moclei of tne calibration data, and use GM(1,1) prediction model to predict the integral developing trend. Then BP neural network is used to build the model of residual, through training to compensate the random fluctuation and get the forecast value of calibration data. GM prediction model based on the residual compensation is given to optimize the calibration interval dynamically, and forecasting model is verified through experiments. Results show that this model has well predicting effect. It can not only predict the integral trend, but also adapt to random fluctuations. Beyond that it is simple, and has strong universality.
出处
《电子测量技术》
2014年第8期56-59,共4页
Electronic Measurement Technology
关键词
残差补偿
灰色模型
BP神经网络
校准间隔
residual compensation
grey model
BP neural network
calibration interval