Método de predicción a corto-plazo de foF2 basado en modelado neuroborroso para su aplicación en sistemas de comunicación por satélite de alta precisión

Abstract

Nowadays, a special attention is being given to the ionosphere influence on the position determination using global navigation satellite system. In this framework, short-term forecasting of ionospheric conditions is gaining a new importance. This work presents a new methodology to predict with 1-24 hours in advance the ionospheric F2-layer critical frequency, foF2. The proposed method is based on artificial intelligence techniques, specifically, on neuro-fuzzy modelling. Neuro-fuzzy techniques have not been extensively used in ionospheric modelling but its application in this field can be efficient and provide successful results. It is well known by scientific community the natural capability that these techniques show to model highly non-linear and complex systems. The method has been tested under quiet and moderately geomagnetic conditions using foF2 data from Slough ionosonde station, providing foF2 forecast (1-24 hours in advance) with relative mean deviation between 4-10%, which is quiet acceptable from practical point of view. A first evaluation of neurofuzzy techniques to model foF2 during severe storm periods has revealed good prediction accuracy for only small (less than 3 hours) lead time prediction. The final purpose will be to check the efficiency of neurofuzzy modelling to predict foF2 with more than 3 hours in advance during disturbed geomagnetic activity periods.