Statistical Sizing of a Satellite Power Subsystem

The Electrical Power Subsystem (EPS) of a satellite is key to its operation in space, as it powers many essential functions for the mission. The EPS must be sized according to the mission power needs, and taking into account the interactions at system level, as it has a high impact on the satellite's mass and cost. The conventional approach for EPS sizing is based on the worst cases of satellite missions and includes several cases of failure. Additional margins are stacked onto the design at interfaces between disciplines and to account for degradation over the satellite's lifetime. It has been observed that this approach systematically leads to an oversized EPS for the intended mission. The aim of this study is to address this overdesign by implementing a new sizing process for the EPS, based on a statistical approach rather than a deterministic one. The approach is implemented here on an existing reference design, proposed for the ESA CRISTAL mission. The proposed approach first focuses on the definition of design parameters and performance metrics for the sizing, and on performing simulations to generate sizing data based on these inputs. Using such data, a sensitivity analysis is conducted to identify the parameters with the highest impact on the design, and surrogate models are trained to efficiently map those parameters to the performance metrics. Uncertainty distributions are then assigned to the design parameters, and propagated to the performance metrics. The obtained distributions are compared to the performance levels that would be achieved with a deterministic approach, highlighting the benefits of the proposed statistical approach. Potential improvements to the approach that shall be considered in future studies are identified and detailed.