一种改进的MAD钟差粗差探测方法

针对传统中位数绝对偏差（median absolute deviation，MAD）方法在探测钟差粗差方面的不足，提出一种改进的MAD钟差粗差探测方法。以一次差分数据为研究对象，通过深入分析传统MAD方法的工作原理，发现传统MAD方法的数学模型表达存在歧义。在优化探测模型结构的基础上，基于岭回归的基本原理，生成动态MAD，有效克服了部分具有显著趋势变化特征的钟差一次差分数据对粗差探测的干扰。使用国际全球卫星导航系统服务组织提供的GPS精密钟差数据进行了实验，并与传统MAD方法进行比较，结果表明，所提方法相较于传统的MAD方法在粗差探测准确率和查全率方面具有明显优势，实际应用价值较高。

An Improved Method for MAD Gross Error Detection of Clock Error

  Objectives  Satellite clock difference (SCB) is affected by many factors, such as data discontinu?ity, jump and gross error, among which the gross error is the most representative. Traditional SCB gross error detection method is the median absolute deviation (MAD) method  Methods  With the single difference of SCB as the research object, we analyzed the working principle of traditional MAD method. In view of the deficiency of traditional MAD in SCB gross error detection, an improved gross error detection method was proposed. Firstly, the ambiguity in mathematical expression was eliminated by optimizing the model structure, and then dynamic MAD was generated based on the basic principle of ridge regression, which effectively overcame the interference of some single difference data with significant trend changes on gross error detection.  Results  We used International Global Navigation Satellite System Service precision SCB as reference for experiments. Compared with traditional MAD, the detection accuracy and recall rate of the proposed method were better than traditional MAD, especially for the data with obvious change of trend characteristics.  Conclusions  The proposed method has obvious advantages in gross error detection accuracy and recall rate and has high application value.