Determinación de un modelo matemático basado en el análisis de un filtro activo para la reducción de distorsiones armónicas mediante Black Box Testing en MATLAB.

Abstract

The present investigation had as objective to determine a parametric mathematical model of an active filter of power for elimination of harmonics distortions produced by an inductive load as continuation of the work developed by the Ing. Lenin Gavilanes who designed and built the system. As an approach to solve the problem in this research, there were applied four identification structures of system, which ones are: the Autoregressive Model with Exogenous Variables (ARX), the Autoregressive of Media Mobile Model (ARMAX), the Output Error Model (OE) and the Box-Jenkins Model (BJ). The system is considered as a black box on where this process of identification is done. First it was developed an acquisition system in Labview to get the input and output system data, then it was taken in account different models with the help of Toolbox Identification System (IDENT) of Matlab, it was based on the view of BEST FIT, LOSS FUNTION and FPE, the analysis of autocorrelation and cross-correlation graphs also was important. In this way to choose the best option for each structure by the application of Cross-Validation and Second Hand Method. Finally it was gotten four models which were exposed for a process of validation with data of second hand different to those used in the stage estimation and validation, at the end to decide that the best parametric mathematical model that is adapted to the system behavior is the estimated based on the structure of Box-Jenkins with coefficient of order [3 2 3 1 0] and with an estimated percentage of 98.87%. With the estimated model is suggested to apply a strategy of control more optimal and easy of implementing to the used in the analyzed filter.