离散历元的动态观测量及其相应的动态模型可能存在异常，若数据处理模型不考虑对这些异常的特别处理，则动态模型参数估值及其所提供的动态信息将极不可靠。基于贝叶斯统计和抗差估计原理，我们构造了一种抗差滤波算法。该算法考虑观测分布和参数验前分布均为污染分布。并利用一个实测网验算该算法和模型的可靠性。 The dynamic observations of discrete epochs and their corresponding dynamic models may be abnormal. If the data processing model does not consider the special treatment of these anomalies, the estimation of the dynamic model parameters and the dynamic information provided by them will be extremely unreliable. Based on Bayesian statistics and anti-error estimation principle, we construct a robust filtering algorithm. The algorithm considers the distribution of observations and the pre-test distribution of parameters as pollution distribution. And use a measured network to verify the reliability of the algorithm and model.