%0 Journal Article
%T 面向卫星物联网的统计业务建模以及实时预测方法
Statistical Traffic Model and Real-Time Prediction Method for Satellite IoT
%A 祁施豪
%A 丁晓进
%A 张更新
%J Software Engineering and Applications
%P 119-133
%@ 2325-2278
%D 2025
%I Hans Publishing
%R 10.12677/sea.2025.142012
%X 针对传统集总业务模型没有考虑终端晶振偏移和电池电量耗尽导致终端死亡，无法反映单个物联网终端状态，同时传统业务模型无法反映单个时隙的准确业务量，需要进行实时预测的问题，我们提出一种面向卫星物联网的统计业务建模以及实时预测方法。首先分析了终端晶振偏移和终端电池耗尽对业务量的影响，其次将传统3GPP业务模型与马可夫泊松过程相结合，考虑影响因素实现了物联网源业务建模，最后提出了一个并行CNN-BiLSTM-Adaboost网络对卫星物联网业务进行预测。仿真结果表明提出的物联网业务建模有效地提升了业务模型的准确性，解决了传统集总业务模型无法反映单个物联网终端状态的问题，同时并行CNN-BiLSTM-Adaboost网络对比传统业务预测方法在预测的准确度上有明显提升。
Aiming at the problem that the traditional aggregated traffic model fails to reflect the state of a single IoT terminal and fails to consider terminal crystal oscillator shift and battery power depletion, and the traditional traffic model fails to reflect the accurate traffic of a single time slot, which requires real-time prediction, we propose a statistical traffic model and real-time prediction method for satellite IoT. Firstly, the influence of terminal crystal shift and terminal battery depletion on the traffic volume is analyzed. Secondly, the traditional 3GPP traffic model is combined with Markov Poisson process to realize the IoT source traffic modeling considering the influencing factors. Finally, a parallel CNN-BiLSTM-Adaboost network is proposed to predict the satellite IoT traffic volume. The simulation results show that the proposed traffic model effectively improves the accuracy of the traffic model, and solves the problem that the traditional aggregated business model cannot reflect the state of a single Internet of Things terminal. Meanwhile, the parallel CNN-BiLSTM-Adaboost network has significantly improved the prediction accuracy compared with the traditional prediction methods.
%K 卫星物联网，
%K 业务模型，
%K 马尔可夫泊松过程，
%K 机器学习
Satellite IoT
%K Traffic Model
%K Markov Poisson Process
%K Machine Learning
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=110095