Qualitative Prediction of Chaotic Behaviors in Time-Varying Lorenz Systems through On-Line Parameter Normalized Identification

JORDAN, MARIO ALBERTO; MURRAY, FEDERICO; PICOLO, CYNTIA

CRIBABB, Bahia Blanca

Conferencia; XXIII Reunión Científica de Geofísica y Geodesia AAGG2006; 2006

IADO, Bahia Blanca

In this work a class of algorithms for estimation of time-varying coefficients of Lorenz-like systems is presented. The main application is aimed at the prediction of weather scenarios providing a solution based on Identification and the Theory of Bifurcation. The estimation algorithms are based on a gradient law and faced the problem of different order of magnitudes in the regression elements and of unbounded time-varying coefficient with the use of normalization. Two mechanisms are developed and compared. Numeric simulations show that qualitative predictions of future behavioral evolution can precisely be made with large anticipation in time.