Terenkovs'kyj I. Multistep prediction algorithms and nonstationary parameter estimation

The dissertation is devoted to development and studying of multistep recursive prediction algorithms and estimation of nonstationary parameters in the presence of disturbances and uncertainties in the parameter drift. Adaptive algorithms using limited member of measurement for error correction were developed. Convergence properties of the algorithms were studied and recommendations for the algorithm parameters (used number of measurements and data weighting coefficient) were proposed. Numerical stability of the proposed algorithms was studied and methods to improve it were proposed. Programming realization, simulation and application of the algorithms to the sodium carbonate production confirmed their effectiveness.