面向未来星地激光骨干链路可靠组网的云预测技术研究

Research on Cloud Prediction Technology for Future Reliable Satellite-Ground Laser Backbone Networking

为解决星地骨干链路组网过程中,激光链路因受到云层干扰而出现的阻塞、中断等影响通信质量的问题,提出了一种基于随机森林算法的云预测技术,训练出高准确率、高提前量的预测模型。预测模型以随机森林为基本算法,与云量相关的气象参数集合为数据集,经过基本模型搭建、两轮超参数优化三个步骤,可以有效提高模型预测准确率。实验结果表明,对有云无云情况的预测准确率达到90%以上,而加入了预测模型的组网系统可以有效降低链路中断次数,提高链路容量。

In order to solve the communication quality problems such as blockage and outage due to the cloud interference to laser links during satellite-ground backbone networking,a cloud prediction technology based on random forest algorithm is proposed,and a prediction model is trained with high accuracy and advance.Using the random forest as the basic algorithm and the cloud meteorological parameters as the dataset,the proposed model can effectively improve the prediction accuracy through the basic model establishment and two-round hyperparameter optimization.The experimental results show that the prediction accuracy in the cloud or cloud-free cases can reach over 90%,and the networking system incorporating the prediction model can effectively reduce the number of link outages and improve link capacity.