Search

An approach to the development of a digital ecosystem (DES) using predictive model and data fusion technics is presented with reference to industrial systems. Predictive models are based on data mining algorithms, wavelet analysis and the creation of knowledge bases. Digital identification models are presented, developed as digital twins for the state of DES. In turn, the control of DES is considered as situational control. The modern approach to the development of architecture for DES is proposed.

The methods for developing predictive models in control systems and decision-making support for nonlinear non-stationary objects are proposed. The methods are based on the application of associative search procedure to virtual model identification as well as Gramian techniques. The associative search methods use intelligent process knowledge analysis. The knowledge base is created and extended in real-time process operation. Intelligent algorithms are offered for predicting power plant dynamics in optimization tasks. Gramian technique of stability analysis for discrete system is used for investigating linear virtual model stability. It is shown that the bilinear Lyapunov equation solutions can be calculated as an infinite sum of the matrix quadratic forms made up by the products of the Faddeev matrices obtained by decomposing of linear subsystem dynamic matrix resolvents.

A conceptual approach to the development of a digital ecosystem using multi-agent technology and predictive models is presented with reference to electric power systems. Predictive models are based on data mining algorithms, wavelet analysis and the formation of knowledge bases.

The paper analyzes the capabilities of identification algorithms based on the intelligent data analysis to predict the approach of process parameters to critical limits, as well as their possible transition to chaotic dynamics. The results obtained are illustrated with an example from the oil refining industry.

Identification models found a wide application in advanced control systems, decision support systems, play a fundamental role in any activity concerned with the signal processing process. Methods based on the wavelet-analysis are characterized by unique possibility of detail frequency analysis in the time to select "frequency-time windows". In the paper, identification methods, developed by the authors, of non-linear time-varying systems by use of the multi-scale wavelet-transform are presented.

Identification models find out a wide application in modern control systems, decision support systems, play a fundamental role in any activity concerned with the signal processing process. Methods based on the wavelet analysis are characterized with a unique possibility to select “fTequency-domain windows”. The paper presents methods of identification of non-linear systems, developed by the authors, by using the multi-scale transformation.