Exploiting Wavelet Recurrent Neural Networks for satellite telemetry data modeling, prediction and control

Multidimensional times series prediction is a challenging task. Only recently the increased data availability has made it possible to tackle with such problems. In this work we devised novel method exploit multiple correlated features in time series. The recurrent neural networks and wavelet transform have been important innovations fields of signal processing prediction. This paper proposes Wavelet Recurrent Network for multi-steps ahead multidimensional proposed model combines these two elements into network that predicts samples future are steps respect input samples. carries out multiresolution decomposition through transform, coefficients transforms output back domain. applied satellite telemetry data, composed readings from sensors which highly correlated. telemetries can help engineers detect anomalies system, that, context space missions, particularly dangerous since they compromise entire mission if not handled properly. results show outperforms without both terms accuracy width forecast horizon. • RNN Transform unique network. Exploits Overcomes traditional analytical numerical models, more expensive. Yields medium-term forecasts an high accuracy. Outperforms simple