动态仿真结果距离检验方法

提出时间序列与其仿真结果之间平均马氏距离的概念,在此基础上建立一种动态仿真结果的距离检验方法.给出正态分布情况下动态仿真结果的拒绝域,提出非正态分布情况下确定拒绝域的两步仿真法.距离检验方法可以根据一个实测样本函数或多个实测样本函数对动态仿真结果进行高置信度的检验,并且对时间序列没有任何限制,即对平稳和非平稳的时间序列均适用.文中还给出正态分布和非正态分布两种情况下时间序列协方差矩阵与其仿真结果的检验方法,以及动态仿真的标准化距离检验方法.

DISTANCE TEST METHOD FOR DYNAMIC SIMULATION RESULTS

A concept of mean Mahalanobis distance between time series and its simulation result is presented. Based on the concept, a distance test method for dynamic simulation results is established. The rejection regions for dynamic simulation results in normal distributions are given, and a two-step simulation method of determining the rejection regions in non-normal distributions is presented. By the method, dynamic simulation results can be tested at a high confidence level according to one measured sample function or multiple measured sample functions. This method has no limitation for time series, namely suitable for the stationary and the non-stationary time series. Furthermore, test method for equality between covariance matrixes of time series and its simulation result is given in the normal and the non-normal distribution respectively. A standardized distance test method for dynamic simulation results is also established.