Forecasting of Time Series Telemetry for Satellite Operations using Deep Learning Techniques

The current expansion of CubeSats being utilized for business and in particular academic ventures has increased in the last decade. This has been positive because businesses can now expand into and offer different services that are only possible within the space sector. CubeSat development requires specialized skills, fiscal commitment, and facilities, which are far less expensive in comparison to vastly larger satellites. However, CubeSats need to be monitored and various parameters need to be adjusted accordingly to fulfil their respective mission requirements. Depending on the available infrastructure and personnel, the analysis of the generated telemetry may often not be addressed until the mission has been completed, and various risks present during the mission would have been mitigated. The purpose of this research is to provide a platform which is a semi-automated deep learning (artificial intelligence) system that will provide an objective assessment of the operational performance of the CubeSat which can be compared against the various resource budgets and/or mission design specifications. In addition, this system will be capable of forecasting telemetry values that the operational team, post-handover, can leverage on as a tool to fine tune the CubeSat operations for their optimum performance. The balance of the study is to have a fully documented developed standalone deep learning system, which with minimal modification may be used in other facilities or systems, such as: where in-depth data analysis for multiple large data sets is needed.