Configuration and development of artificial neural network model for spacecraft power supply system control system under the conditions of uncertain factors

The paper considers uncertain factors that can lead to abnormal situations in the control system of the power supply system of a spacecraft. Certain factors that can be predicted as well as factors whose influence can be accounted for when designing the control system and building control algorithms are highlighted. Uncertain factors that can be predicted using the intellectualization of electric power distribution control system have been identified. Elements of the system the reliability of which can be improved by applying intelligent control system and the prediction of abnormal situations on the basis of artificial neural networks have been identified. The analysis of existing control algorithm for power supply system has been carried out. By means of the telemetry parameters used in this algorithm, selected telemetry parameters for use in the intelligent control system of the power supply system have been identified. The criterion for an emergency situation the occurrence of which must predict the artificial neural network is defined. The configurations of artificial neural networks which can be used as a foundation for intelligent control system of power supply system of a spacecraft are considered. The problem of available training data sample optimization for training the artificial neural network is regarded. Suitable methods for the optimization of neural network training in the context of the specifics of the problem are considered. A specific configuration of artificial neural network, mindful of the specifics of application and the heterogeneous nature of the training data sample, is proposed.

1. Alekseev V.P., Kovalev A.P. Factors determining the reliability and durability of spacecraft avionics structures. Novye issledovaniya v razrabotke tekhniki i tekhnologii. 2015;2:24–28. (In Russ.).

2. Kuznetsov N.V., Panasyuk M.I. Space Radiation and Fault and Fault Tolerance Prediction of Integrated Circuits onboard Spacecraft Equipment. Voprosy atomnoi nauki i tekhniki (VANT), Seriya “Radiatsionnoe vozdeistvie na radioelektronnuyu apparaturu” = Questions of atomic science and technics. Series: Physics of radiation effects on radio-electronic equipment. 2001;1–2:3–8. (In Russ.).

3. Savenkov V.V., Tishchenko A.K., Volokitin V.N. Principles of construction of regulation and control equipment of modern power supply systems for small spacecrafts. Reshetnevskie chteniya. 2017;21(1):325–326. (In Russ.).

4. Brink H., Richards J.W., Fetherolf M. Real world Machine Learning. US: Manning Publications Co; 2017. 266 p.

5. Flakh P. Machine Learning. The science and art of building algorithms that extract knowledge from data. translated from English by A.A. Slinkin. Moscow: DMK Press; 2015. 400 p. (In Russ.).

6. Zhanaeva S.B. To the question of data preparation in the development of a neural network model. Vestnik SibGUTI = The Herald of the Siberian State University of Telecommunications and Information Science. 2022;4(60):69–79. (In Russ.).

7. Sholle F. Deep Learning in Python. Programmer's Library Series. Sankt-Petersburg: Piter; 2018. 400p. (In Russ.).

8. Krasnov S.S., Kuralesova N.O. Selection of a neural network model for a decision-making system for controlling complex technical devices. Vestnik Volzhskogo universiteta im. V.N. Tatishcheva = Vestnik of Volzhsky University after V.N. Tatischev. 2013;4(22):92–96. (In Russ.).

9. Kashirina I.L., Demchenko M.V. Research and comparative analysis of optimization methods used in the training of neural networks. Vestnik Voronezhskogo gosudarstvennogo universiteta = Proceedings of Voronezh State University. 2018;4:123 132. (In Russ.).

10. Wilson A.C., Roelofs R., Stern M., Srebro N., Recht B. The Marginal Value of Adaptive Gradient Methods in Machine Learning. US: Cornell University Library; 2017. 14 p.