基于卫星遥感监测极端气象预报数据异常值检测方法

Detection method for outliers in extreme weather forecast data based on satellite remote sensing monitoring

摘要:

在遥感数据采集过程中,由于传感器故障、气象条件等原因,可能会导致少量的异常点出现在采集的数据中,这些异常点可能会对极端天气预报的准确性产生负面影响。为此,需要研究一种基于卫星遥感监测极端气象预报数据异常值检测方法。基于改进K-均值聚类算法对缺失的卫星遥感监测极端气象预报数据进行插补,还原数据完整性。划分星遥感监测极端气象预报数据区段,提取每个区段的四个特征参数,以此为输入,利用蝙蝠算法优化BP神经网络识别异常区段。计算异常区段中每个卫星遥感监测极端气象预报数据的局部离群因子,局部离群因子大于1.0数据为气象预报数据异常值,以此完成气象预报数据异常值检测。结果表明:所提方法插补误差小于±1.0,可以准确识别异常区段中的异常值,且在不同样本中的协调指数高于0.8,检测效果更好。

Abstract:

In the process of remote sensing data collection, due to sensor failures, meteorological conditions, and other reasons, a small number of abnormal points may appear in the collected data, which may have a negative impact on the accuracy of extreme weather forecasting. Therefore, it is necessary to study a method for detecting outliers in extreme weather forecast data based on satellite remote sensing monitoring. Based on the improved K-means clustering algorithm, the missing satellite remote sensing monitoring extreme weather forecast data is interpolated to restore data integrity. Divide extreme weather forecast data sections for satellite remote sensing monitoring, extract four feature parameters for each section, and use them as inputs to optimize BP neural network recognition of abnormal sections using bat algorithm. Calculate the local outlier factor of extreme weather forecast data monitored by each satellite remote sensing in the abnormal section. Data with a local outlier factor greater than 1.0 are considered abnormal values of weather forecast data, in order to complete the detection of abnormal values of weather forecast data. The results show that the interpolation error of the proposed method is less than ± 1.0, which can accurately identify outliers in the abnormal section. Moreover, the coordination index in different samples is higher than 0.8, and the detection effect is better.