Telemetry Prediction within the Mission Planning System: Optimizing the Battery Unitization of the TerraSAR-X and TanDEM-X Satellites

A novel approach on the decision-making of the Mission Planning System is developed for the TerraSAR-X and TanDEM-X satellites. Combining the available data from the ground and the space segment, a continuous quantitative study is performed on the batteries of the TSX and TDX satellites. The first model developed as an outcome of this study is the chain model that groups the SAR acquisitions of the satellites depending on their time separation. Processing a long period of past data allowed us to estimate accurately the battery performance in upcoming periods, especially around the summer solstice of the following year, where the poorest battery performance is expected. Evolving on the estimation concept, we developed a new model that estimates the battery voltage telemetry for every SAR acquisition during the planning run. Depending on the level of the predicted telemetry parameters values, the Mission Planning System can decide if an activity should be planned or not. Once real telemetry is available on ground, the status of the batteries is compared to the estimated values. The inevitable ageing of the satellites is the main driver for a continuous study on their batteries. Our objective is a) to preserve the good performance of the batteries, supporting all the mission operations, while b) prevent any event on the batteries that might interrupt the operations. As a result in the last years, as long as the machine learning algorithms are running in the Mission Planning System, we filter all the on-board activities in which the battery-related telemetry would have exceeded the nominal limits, while we maintain the activities volume at the same levels as before, succeeding the optimization of the battery utilization. Applying this logic to the battery telemetry parameters is proving the concept of predicting accurately the telemetry during the planning run. This does not only lead to off-limits-violation prevention during the generation of the timeline, but also to the detection of unexpected onboard anomalies. This approach could potentially be applied on several other satellite units, in order to optimize their utilization via machine learning, regression, algorithms