CubeSat requirements in terms of size, weight, and power restrict the possibility of having redundant systems. Consequently, telemetry data are the primary way to verify the status of the satellites in operation. The monitoring and interpretation of telemetry parameters relies on the operator’s experience. Therefore, telemetry data analysis is less reliable, considering the data’s complexity. This paper presents a Machine Learning (ML) approach to detecting anomalies in solar panel systems. The main challenge inherited from CubeSat is its capability to perform onboard inference of the ML model. Nowadays, several simple yet powerful ML algorithms for performing anomaly detection are available. This study investigates five ML algorithm candidates, considering classification score, execution time, model size, and power consumption in a constrained computational environment. The pre-processing stage introduces the windowed averaging technique besides standardization and principal component analysis. Furthermore, the paper features the background, bus system, and initial operational data of BIRDS-4, a constellation made of three 1U CubeSats released from the International Space Station in March 2021, with a ML model proposal for future satellite missions.

This article presents a database containing on-orbit data samples of the Electrical Power System (EPS) from 4 different 1U CubeSats belonging to the BIRDS constellation. The EPS is responsible for providing uninterrupted power to overall satellite both during sunlight and eclipse. The satellites are based on the BIRDS open-source standardized bus designed by Kyutech for research and education. BIRDS bus was used for six satellites that were delivered to ISS on board the Cygnus re-supply spacecraft launched by Antares rocket and released from International Space Station (ISS) into ISS orbit (altitude 400 km, inclination: 51.6°, duration: 92.6 min). The dataset contains the data of voltage (mV), current (mA) and temperature (Celsius) of the battery and solar panels attached to 5 sides of the satellite. This data is collected by the on-board computer every 90 seconds in nominal operation or every 10 seconds in fast sampling mode. The data is downloaded from the satellite memory by the ground station operators. Next, space engineering experts from Kyushu Institute of Technology have analysed the dataset to classify each data sample into normal or anomaly classes. This paper provides one datafile per satellite, that includes data from solar panels and battery since their deployment into orbit until the end of its life for the UGUISU, RAAVANA, and NEPALISAT satellites, first two showing a failure in one of their panels during more than two years of operation on-orbit. The TSURU satellite dataset includes data since its deployment into orbit and will continue to be collected until the end of its life. The dataset generated will be useful for 1U CubeSat, such as BIRDS platform, users, and satellite developers by using it as a reference to compare the behaviour of their Electric Power System under different operating scenarios and align their missions according to the available power on-orbit. At the same time, the dataset can help computer science researchers to build and validate new models for fault diagnosis and outlier detection.