FILTRATION AND PREDICTION QUALITY COMPARATIVE ANALYSIS OF TWO- AND THREE-LOOP ADAPTIVE EXPONENTIAL FILTERS

Abstract

Introduction. The article provides a comparative analysis of filtration and prediction quality of the adaptive exponential two-loop and three-loop filters. The main difference between two- and three-loop filters is the number of filtering loops that are used to evaluate the quality of the filtration process, as well as their software implementation. Aims. To consider the feasibility of using a three-loop filter-predictor in control systems, as a data processing algorithm, in comparison with two-loop filter. Methodology. I applied the concepts of time series analysis and mathematical modeling in the Matlab. Results. I obtained the characteristics of filtration and prediction standard errors depending on the number of steps which the forecast is made for and the quantity of steps which the data processing algorithm uses for quality estimation of filtering process for two filters variations. Originality. For the first time I have defined the relationships between rms-errors (filtration and forecast) and the following parameters: the number of steps which the forecast is made for, the quantity of steps which the data processing algorithm uses for quality estimation of filtering process for two- and three-loop filtering and prediction algorithms. I analyzed the relevance of applying two different smoothing coefficient adaptation algorithms depending on computer resources. Practical value. I have programed the proposed algorithms using Matlab programming language. This software can be integrated into various automated control systems in order to filter and predict the values of the signal distorted by noise. This research provides an opportunity to choose an efficient data processing algorithm, depending on the task.