Excess Attenuation Prediction At Q-Band Using Deep Learning for High-Throughput Satellite Systems

The increasing demand for higher data rates to support the numerous multimedia services using satellites has led to the employment of much higher frequencies bands above 10 GHz. In these frequencies, the electromagnetic wave attenuation due to rainfall (the rain attenuation) is the most noticeable component of the excess losses. This equals to the satellite channel gain for line of sight (LOS) satellite links. The prediction of this factor is of crucial importance for the satellite system design and the accurate estimation of outage events and for the application of fade mitigation techniques. However, the adoption of data-driven models is hampered by the absence of reliable satellite propagation measurements in real conditions. Here, we exploit real propagation measurements from the two experimental stations in Greece at Q-band to predict the the induced of in excess attenuation values for the next future time-steps. The proposed model is a temporal sequential deep learning scheme with causal convolutions. The results are very encouraging since the sequential model achieves performance 0.3 dB, whereas the respective temporal models in literature achieve performance equal to 0.5 dB for the testset in our dataset.