基于随机森林的卫星快变遥测数据建模

现代卫星已逐渐成为国家重大基础设施，为了解其在轨运行状态，需要对遥测数据进行分析；其中快变遥测数据包含了大量卫星服务情况信息，对该数据进行基于机器学习算法的分析建模，可以更好利用特征维度高、数据量大的快变遥测数据，为人工智能在卫星数据建模、运维方面提供一种可能方案；提出一种基于随机森林算法对在轨卫星快变遥测数据进行建模的方法，并引入改进的二次网格搜索方法对模型参数进行调优；使用模型对某频点功率测量值进行预测，结果显示R2值达到0.98以上，预测值误差较小，建立了效果较好的快变遥测数据模型，为实现基于机器学习的快变遥测数据分析提供了一种可能的方案；

Modeling of Fast-changing Telemetry Data of Satellite Based on Random Forest

Modern satellites have gradually become a major national infrastructure. In order to understand satellite on orbit working status, it is necessary to analyze the telemetry data. The fast-changing telemetry data contains a large amount of satellite service information. The analysis and modeling of the data based on machine learning algorithm can make better use of the fast-changing telemetry data with high feature dimension and large amount data, and provide a possible scheme for artificial intelligence in satellite modeling, operation and maintenance. A method of modeling the fast-changing telemetry data of on orbit satellite based on random forest algorithm is proposed, and an improved dual grid search method is introduced to optimize the model parameters. The model is used to predict the power measurement value of a frequency point. The results show that the R2 value is more than 0.98 and the error of the prediction value is small. A fast-changing telemetry data model with good effect is established, which provides a possible scheme for the analysis of fast-changing telemetry data based on machine learning.