基于时间序列和聚类的挤压机能耗异常检测研究

Study on Abnormal Energy Consumption Detection of Extruder Based on Time Series and Clustering

挤压机能耗数据异常分为点异常和模式异常,传统的异常检测方法往往只能检测出其中一种。基于此,使用一种基于时间序列和聚类的能耗异常检测方法。针对点异常,提出一种基于K-Means和LOF的算法(K-MLS);而对于模式异常,设计K-异常因子(K-MLOF)检测算法。为验证方法的有效性,利用某铝型材企业能耗数据对其进行验证和对比分析,结果表明该方法具有较高的精度。

The abnormal energy consumption data of extruder can be divided into point anomaly and pattern anomaly, and the traditional anomaly detection method can only detect one of them. Therefore, this paper uses a method based on time series and clustering to detect abnormal energy consumption. Aiming at the point anomaly, an algorithm based on K-MeanS and LOF was proposed（K-MLS）. For pattern anomaly, a K-nearest neighbor anomaly factor detection algorithm （K-MLOF）was designed. In order to verify the validity of the method, the energy consumption data of an aluminum profile enterprise were used to verify and compare the results. The results show that the method has higher accuracy.