Methods for the Invalidation/Validation of Simulation Models

Abstract

In the context of simulation model development, the word “validation” is used to describe the process of establishing the extent to which a given model is an accurate representation of the corresponding real-world system. In practice, the emphasis should always be on “invalidation” since a model can never be proved to be, in any general sense, “correct”. This chapter provides an introduction to methods to be used for this process and includes quantitative methods, such as those based on system identification and parameter-estimation principles, barrier certificate methods, techniques based on concepts of model distortion and methods based on parameter sensitivity analysis. Methods of face validation, involving a more subjective approach and the opinion of experts, familiar with the real system, are also considered. That type of validation process is illustrated using an example of the development and testing of a complex simulation of a hydro-turbine generating system. Further sections of this chapter are devoted to approaches based on comparison with other models, the choice of data sets for model testing and validation, the validation of sub-models and generic models, issues relating to the validation of distributed parameter models and the validation of discrete-event and hybrid models. Before the final discussion section there is also a short review of issues arising in the acceptance or upgrading of models.