The Construction of Simplified Models of a Multi-Timescale System on the Basis of Processes Implementations

摘要

The report is devoted to the problem of constructing simplified models of multi-timescale dynamic systems. The general case of decomposition of a two-timescale linear system model with the reduction of its dimension by several orders is considered. A feature of the decomposition procedure is the use of processes implementations corresponding to state variables as initial information. This makes it possible to obtain implementations experimentally as well as analytically in the presence of a description of the system in the Cauchy normal form. We propose a procedure for finding holonomic relationships between the variables that make it possible to reduce the order of the initial model, as well as methods of residual functions or sliding determinant to determine the restrictions on the simplified model time domain. The design estimate of accuracy of the simplified model is given, followed by refinement by simulation of processes. An example illustrating the proposed approach to constructing a simplified model of a multi-timescale system is given. The procedure for constructing a simplified model can be fully implemented by means of MATLAB/Simulink software package.