钢铁企业高炉煤气发生量异常数据检测

从高炉煤气生产的实际工况出发,对异常数据产生的原因和特点进行分析。针对现有异常检测方法运算效率低下的问题,提出一种改进的局部异常因子检测算法。该算法首先利用五数总括法剔除掉大量的正常数据,然后再用一种相对k距离的比值表示剩余离群点的异常程度,进而判断异常数据。仿真实验表明:改进方法检测所需的时间比传统的局部异常因子方法检测所需的时间更少,且检测效果更加准确、直观。

Outliers Detection of BFG Generating Data in Iron and Steel Industry

Starting with actual conditions of blast furnace gas ( BFG) generation, both causes and characteris-tics of outliers in BFG generation data were analyzed; and considering poor operation efficiency of existing methods for the outliers detection, an improved outlier factor detection algorithm was proposed, in which, hav-ing the five-number method used to eliminate a large number of normal data and then having the method of rel-ative k distance’ s ratio adopted to express the degree of abnormity of the rest data and then to determine the outlier. The simulation experiment shows that this improved outlier factor algorithm becomes more time-saving and accurate than traditional local outlier factor algorithm.