基于改进型BP神经网络的氢原子钟钟差预测

原子钟的钟差预测是原子钟时标计算和原子钟驾驭的关键环节,良好的钟差预测可明显提高原子钟时标和原子钟驾驭的精度。为进一步提高氢原子钟的钟差预测精度,本文提出了一种改进型的BP神经网络算法,并用中国计量科学研究院守时实验室氢原子钟组的实际数据进行了验证。验证结果表明,本文提出的改进型BP神经网络钟差预测算法与目前采用的线性回归钟差预测算法、SVM钟差预测算法相比,显著地提高了氢原子钟钟差预测精度。该钟差预测算法的提出对提高原子钟时标和驾驭精度有很好的推动作用。

Hydrogen atomic clock difference prediction based on improved back propagation neural network

Atomic clock difference prediction is the key step of atomic clock time scale calculation and atomic clock control. Good prediction of the clock difference can significantly improve the precision of the atomic clock time scale and atomic clock control. In order to further improve the prediction precision of the clock difference of the hydrogen atomic clock, this paper presents an improved BP neural network algorithm to predict the atomic clock difference, which is verified by the actual hydrogen atomic clock data of the time-keeping laboratory in National Institute of Metrology, China. The verification results show that compared with the clock difference prediction algorithms based on linear regression and SVM used at present, the improved BP neural network clock difference prediction algorithm significantly improves the prediction precision of hydrogen atomic clock, and has a good promoting effect on improving the precision of atomic clock time scale calculation and atomic clock control.