基于集成学习的空间科学卫星工作模式识别

Recognition of Working Pattern of Space Science Satellite Based on Ensemble Learning

针对空间科学卫星遥测参数数据量大且特征维度高、需要消耗大量人力资源预先设置海量阈值、预先设置的阈值可能不再适用、现有监测手段可扩展性低等问题,提出了一种基于集成学习的空间科学卫星工作模式识别方法。该方法采用相关系数统计特性和互信息理论对遥测参数数据进行筛选降维,使用数据重采样技术解决数据集中存在的类别不平衡问题,构建集成学习模型,实现空间科学卫星工作模式的识别。借助某型号科学卫星真实遥测参数数据对该方法进行验证,在短时内便可构建完成算法模型,模型对整体类别的识别正确率高达99.67%,可正确识别多数类样本和少数类样本,为地面运控人员判断空间科学卫星工作模式提供了决策依据。

Aiming at the issues of space science satellite telemetry parameters,such as large amount of data,high dimension,the need of numerous artificial resource consumption for preset massive thresholds,the preset thresholds that may not be applicable,and the current monitoring methods with low scalability,a working pattern recognition method is proposed for scientific satellite based on ensemble learning.Correlation coefficient statistical characteristics and mutual information theory are used to screen and reduce the dimension of telemetry parameter data.Data resampling technology is used to solve the problem of category imbalance for the dataset.An integrated learning model is used to identify the working mode of space science satellite.The method is verified with the real telemetry parameter data of quantum science satellites.And the algorithm model can be constructed in a short time,and the overall recognition accuracy rate reaches 99.67%,which can correctly identify the majority and minority class samples.The method can provide decision-making basis for ground personnel to judge the working mode of space science satellites.