Abstract:

To validate the dynamic output of the simulation system under uncertainty, a model validation method based on the data feature is introduced. For the purpose of measuring the difference of the simulation and reference output comprehensively, three kinds of feature extraction models including shape, position and frequency spectrum are given to describe the difference feature of the simulation and reference output. Standard reference series is set up and the feature validation matrix is built to measure the uncertainty of the outputs. The Bayesian factor method improved by kernel density estimation is applied to validate the consistency of simulation dynamic outputs based on the feature validation matrix. Finally, the method is validated in the application.