IMPROVEMENT OF QUALITY OF THE EVALUATION OF MODEL PARAMETERS BY INTEGRATED METHOD OF LEAST SQUARES

Abstract

Modern systems of management, optimization and forecasting of production require new algorithms of work and evaluation of systems and processes. When solving problems, in the absence of a priori information, there is a need for the use of effective methods of parametric identification. The accuracy of the existing classical methods of identification depends on the availability of certain information regarding the characteristics of the signals, in particular, on the law of the distribution of random error of measurement, and therefore in the real processes are ineffective. There is a need to apply methods that provide more accurate estimates of the parameters of a mathematical model of the object under study in time-limited and non-sensitive data, about noise variables and control impacts. The estimation of parameters is carried out using the integrated method of least squares, which provides smoothing of the external influences of the model under study on the results. The effectiveness of the method under consideration is confirmed by comparison with the least squares method. The optimization of the parameters of the weight function according to the external criterion has been made, at least the norm of the difference between the estimates of the parameters of the pair and odd sequences. The analysis of the dependence of the accuracy of the estimations of parameters on the choice of coefficients of the weight function is carried out.