基于Map/Reduce的电力监控系统规则挖掘技术研究

Power Monitoring System Rule Mining Research Based on Map/Reduce

鉴于电力监控系统的多异类信源和动态融合特性,对系统海量实时监控信息进行故障规则挖掘和预测性监测。针对传统决策树法效率难提升问题,提出一种基于并行框架Map/Reduce和包含度测量相融合的规则挖掘算法(MRDT)。通过构建云计算Hadoop平台,在其分布式并行计算框架Map/Reduce基础上实现基于包含度的决策树规则挖掘算法的并行处理,高效地提取信任度较高的故障规则。以某水电站实时监控系统的电气信息为例,对MRDT算法进行实验测试,结果表明:MRDT算法在保证传统DT算法规则信任度较高的同时,提高了挖掘效率。

At present, multiple heterogeneous sources and dynamic fusion characteristics of the power monitoring system become more obvious, fault rule mining and predictive monitoring methods could be applied to large real-time monitoring information, which could reduce economic loss. In view of the lower efficiency problem of the traditional decision tree method, a mining algorithm（MRDT） based on parallel framework Map/Reduce and contains fused measurement rules is put forward. Through the construction of Hadoop cloud platform, parallel processing of decision tree rule mining algorithm based on inclusion degree is completed efficiently on Map/Reduce framework. Taking the electrical information of the hydropower station monitoring system as an example, experimental results show that the MRDT algorithm ensures the high trust of rules of the traditional DT algorithm, while obviously improves the mining processing efficiency and provides a more efficient processing method for exception prediction of the power monitoring system.