The thesis is concerned with creating a basis for the design of robust controllers, which pertain, in particular, to power station processes. The basis is carried with the help of a known two staged identification method. The purpose of the identification method is to provide, in addition to the usual nominal model, a set of interval models which represents the uncertainties of the identified models. This should be a pertinent basis for designing robust controllers under uncertainty conditions.

The main idea is identification of SISO models family using a number of data subsets, which are selected from the whole data set by a special procedure. According to predefined criteria nominal and uncertainty models are calculated from this models family. For exact calculation of maximal frequency range, in which a model correctly describes the process, a model transfer function is represented by transfer function with interval coefficients. A possible method for calculation of such interval coefficients is proposed. The maximal frequency range test is based on calculating frequency response envelopes for such ”interval” model by the analytical method [26]. The method of calculation of common maximal frequency range (where both nominal and uncertainty models correctly describe the process) is proposed.

In addition, the problem of improving identification accuracy for a differentiating process is solved. Adaptive prefiltering of data is considered as a possible solution. Suitable scheme for searching variable filter parameter is presented by using the proposed criterion. Main theoretical results are illustrated by practical applications at the Israel Electric Corporation.