Scholarlink

<p>Geostationary orbit (GEO) communication satellites are equipped with several temperature sensors on their antennas, and the data collected from these sensors are utilized for various operational purposes. By virtually modeling these sensors, reducing the number of physical sensors and associated system complexity becomes feasible, thereby lowering overall costs. This study aims to model the thermal sensor of a GEO communication satellite antenna using artificial neural networks (ANN) and develops a virtual thermal sensor based on actual telemetry data. The dataset used spans 92&#xa0;days, covering the months of May, June, and July. The performance of the developed virtual thermal sensor was evaluated by comparing the predicted values with actual measurements and analyzing the results using statistical methods. The proposed thermal sensor model achieved a regression (R<sup>2</sup>) value of approximately 0.999 and a mean square error (MSE) of around 3 × 10⁻<sup>3</sup>, highlighting its high accuracy. Specifically, predictions yielded an average MSE of 1.958 × 10⁻<sup>3</sup> across multiple test scenarios, including both short-term (single day) and long-term (three day) periods. These numerical results clearly indicate that the ANN-based virtual thermal sensor model effectively predicts GEO satellite antenna surface temperatures, providing a reliable alternative in sensor failure scenarios or reducing sensor complexity.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Predictive modeling of satellite antenna temperature sensors using artificial neural network

  • 期刊论文

  • 发表日期：

    2025年9月3日

查看全文
  • Ibrahim OZ

摘要

Geostationary orbit (GEO) communication satellites are equipped with several temperature sensors on their antennas, and the data collected from these sensors are utilized for various operational purposes. By virtually modeling these sensors, reducing the number of physical sensors and associated system complexity becomes feasible, thereby lowering overall costs. This study aims to model the thermal sensor of a GEO communication satellite antenna using artificial neural networks (ANN) and develops a virtual thermal sensor based on actual telemetry data. The dataset used spans 92 days, covering the months of May, June, and July. The performance of the developed virtual thermal sensor was evaluated by comparing the predicted values with actual measurements and analyzing the results using statistical methods. The proposed thermal sensor model achieved a regression (R2) value of approximately 0.999 and a mean square error (MSE) of around 3 × 10⁻3, highlighting its high accuracy. Specifically, predictions yielded an average MSE of 1.958 × 10⁻3 across multiple test scenarios, including both short-term (single day) and long-term (three day) periods. These numerical results clearly indicate that the ANN-based virtual thermal sensor model effectively predicts GEO satellite antenna surface temperatures, providing a reliable alternative in sensor failure scenarios or reducing sensor complexity.