The Application of Machine learning to Amazonia-1 satellite power subsystem telemetry prediction

Abstract This article presents the data acquisition, exploratory analysis, model training, evaluation, and use of hyperparameters in a machine learning that will be used to predict telemetry from Amazonia-1 satellite. The satellite was launched 2021, it uses Multi-Mission Platform as service module has Wide Field Imager imaging camera. Its power subsystem 715 telemetries with distinct types dependent independent variables. amount generated daily is large, making manual analysis this unfeasible. ensemble XGBoost algorithm values variable D008 “Battery Module 1 Voltage” belongs electric subsystem. For evaluation performance Mean Absolute Error (MAE), Root Square (RMSE), R2 are used. final resulted coefficient determination (R 2 ) 99.99%, MAE 0.005749, RMSE 0.007727. After cross-validation step, reached 0.006888. execution time 57 minutes 32 seconds. Based on these numbers, we can consider built good result, especially when cross-validation.