基于DBSCAN聚类的热能发电大数据异常检测模型

Anomaly Detection Model of Thermal Power Generation Big Data Based on DBSCAN Clustering

为了解决热能发电大数据异常检测时,存在误报率高、检测率低和漏报率高的问题,提出了基于DBSCAN聚类的热能发电大数据异常检测模型。首先通过时间序列模型对原始发电大数据进行异常值修正,然后将修正后的数据归一化处理,最后基于Spark Streaming设计Streaming DBSCAN算法,结合历史数据和相似发电厂数据的聚类特征,完成热能发电异常数据的检测。实验结果表明,所提方法可以有效降低误报率和漏报率、提高检测率并准确地获取异常值。

In order to solve the problems of high false positive rate, low detection rate and high false negative rate in thermal power generation big data anomaly detection, a thermal power generation big data anomaly detection model based on DBSCAN clustering is proposed. Firstly, the outliers of the original power generation big data are corrected through the time series model, and then the corrected data are normalized. Finally, the streaming DBSCAN algorithm is designed based on spark streaming. Combined with the clustering characteristics of historical data and similar power plant data, the detection of abnormal data of thermal power generation is completed. Experimental results show that the proposed method can effectively reduce the false positive rate and false negative rate, improve the detection rate and accurately obtain abnormal values.