Data stream mining for predicting the thermal power consumption of the Mars Express spacecraft

Abstract

The Mars Express (MEX) spacecraft, operated by the European Space Agency (ESA), has been orbiting Mars for the past 18 years. During this period, it has provided unprecedented scientific data about the red planet, but it has also aged, and its batteries have degraded. Thus, MEX needs careful and accurate power modeling to continue its significant contribution without breaking, twisting, deforming, or failure of any equipment. The power consumed by the autonomous thermal subsystem, that keeps all equipment within its operating temperature in a difficult environment, is the only unknown variable in the spacecraft’s power budget. In this pilot study, we address the task of predicting the thermal power consumption (TPC) of MEX on all of its 33 thermal power lines, learning predictive models from the stream of its telemetry data, which is a task of multi-target regression on data streams. To analyze such data streams and to model the MEX power consumption, we consider both local and global approaches, i.e., predicting each target by a separate model and predicting all targets at once by a single model, respectively. Our evaluation of the considered approaches investigates their performance in predicting the MEX power consumption, the influence of the time resolution of the measurements of TPC on this performance, and the success of the methods in detecting and adapting to change.