Running Simulation Models in Parallel With Physical Systems for Improved Estimation Performance: Semantic Models Facilitate Updating Model State, Parameters, and Structure

When a simulation model for a dynamic system is used in parallel with the real physical system adaptation is a very important design feature. Model adaptation allows the model to compensate for errors between the simulated and real outputs. While parameter tuning and state updates can be automated to improve model accuracy, automatically updating the structure of a model is generally difficult to do. In order to build an automatic structural adaptation scheme we propose a semantic architecture for the model definition and use the semantic form to design an adaptation law. The semantic structure provides the standardization required for an automated algorithm to derive meaning from the model definition while allowing model designers to retain some generality in the way models are built. The semantic modeling approach is presented with an example implementation and discussion about its application to wide area distributed control systems. The wide area control system we use as motivation is a regional electrical power distribution controller.