Recursive identification of inertia tensor parameters of space solar power satellite based on distributed placement of attitude sensors

Abstract

This study addresses the issue of recursive identification of the inertia tensor parameters for a space solar power satellite with planar configuration based on a distributed placement of its attitude sensors. Different from the centralized configuration of the attitude sensors in a traditional rigid spacecraft, multiple attitude sensors are used to reduce the impact of the vibration signals of the large flexible structure on the measured attitude signals and improve the identification accuracy. By establishing an attitude–vibration dynamic model of the system, achieving optimal placement of multiple attitude sensors is investigated and a corresponding optimization criterion is developed. Subsequently, based on the least squares theory, the variable forgetting factor is derived by minimizing the system mean square deviation. Moreover, an improved recursive least squares method is developed to estimate the system inertia tensor parameters using the coupled least squares estimation. Numerical simulations are conducted, and based on the attitude output signals obtained from the optimal sensor placement results, the inertia tensor parameters are identified using the developed algorithm. The results demonstrate that the structural flexibility significantly affects the identification accuracy of the inertia tensor parameters, and the distributed placement of the multiple attitude sensors reduces this influence to a certain extent. Compared to a single group of attitude sensors, the ten groups of sensors for identifying the inertia tensor parameters reduce the absolute and relative values of the average relative error by approximately 3.3%–5.4% and 36.5%–55.6%, respectively. In addition, the identification results reveal that the absolute value of the average relative error is decreased by approximately 2.8%–3.5% on using the proposed recursive algorithm compared to that obtained using the conventional recursive least squares method when the signal-to-noise-ratio is 10 dB.