Optimization of Identification of Non-Stationary Objects Based on the Regulation of Systematic Error Values

The problem is formulated and methods are developed to increase the reliability of information based on the synthesis of methods of statistical, dynamic, fuzzy identification, threshold control, control by increments and with prediction, assessing the influence of factors on the systematic error and mechanisms for adjusting model parameters to optimize data processing of non-stationary objects. Methods of multivariate analysis are proposed to improve the efficiency of identification and approximation of objects, representing data in the form of random time series. The developed methods are recommended for implementation in the form of software-algorithmic complexes to ensure the credibility of data processing based on hybrid identification, taking into account the nonlinearity of influencing factors and the uncertainty of parameters in dynamic models.