Earth-Space Rain Attenuation prediction using Optimum Algorithm of Artificial Neural Networks

Multiple applications and services offered through different types of satellites resulted in extreme congestion of low frequency bands. Due to this, exploitation of high frequency spectrum is chosen in allocating new services as a response to high demands. Satellite networks operating on those high frequency spectrums of microwaves and millimeter waves suffers from the effect of rain on signal propagation through earth-space links. The rain attenuation effect is noticeable from as low as 10 GHz depending on climate conditions of geographical location. Various rain attenuation prediction models are being developed, modified and adopted to improve prediction with the aim of establishing and maintaining the reliable satellite networks around the world. This paper presents an improved estimation of rain attenuation on satellite communication links using Artificial Neural Networks (ANNs) in four provinces of South Africa. The paper explore feedforward back-propagation (FFBP), cascade forward back-propagation (CFBP), feedforward distributed time-delay (FFDTD) and Elman backpropagation (EBP) as learning algorithms. The results show that neural networks models produced a relatively less error as compared to International Telecommunication Union model and Simple Attenuation Model.