摘要
通过基于"厂商+批次"对电能表整体运行状态进行分析,不仅可以发现电能表的运行故障率呈现出明显的层次分布,同时还能发现家族性的问题或者缺陷,实现基于传统的人工经验诊断转变为基于机器学习智能分析预测。第一阶段:以厂商和生产批次为对象,通过对电能表状态的故障率、报废费和折旧率进行分析,将所有电能表的分析数据降维整合为"非健康度曲线"的一维数据,且利用散点图将分析对象非健康值展现。不仅能告诉我们每个批次电能表的现状,还能告诉我们哪些批次存在问题,根据不同的预警等级,确定电能表故障的严重性。第二阶段:通过对电能表工作状态和工作环境实时监测,借助机器学习中线性回归的算法,诊断、预测电能表的实际运行状态,预测电能表非健康度值变化趋势。基于上述二个阶段的分析,为电能表状态检修、备品备件等工作提供辅助决策依据。
Through the comprehensive analysis of the overall running status of energy meters based on ＂Manufacturer plus Batch＂ pattern, we can find that the operation fault ratio of the energy meters has obvious layered distribution, and the problems or defects have family characteristics. This analysis helps to change the diagnostic method based on the traditional human-based experience into the machine-based learning inte- lligent analysis and forecasting. The first stage： with manufac- turer and production batch as the object, the fault rates, scrap costs and depreciation rate of the energy meter status are analyzed and then all the analysis data of the energy meters are integrated into one-dimension data through the dimension reduction process, and the scatter plot is used for the non- health value show for the analysis object. This process not only shows the current status of each batch of energy meters, but also tells us which batch （or batches） has problems and identify seriousness of the meter fault by different warning degrees. The second stage： though the real-time monitoring of the working status and the working environment of the energy meter, the actual running state of the meter is diagnosed and predicted by the linear regression algorithm in the machine learning, and the change trend of the non health value of the electric energy meter is also predicted. Based on the analysis of the two stages, this paper can provide certain supplementary decision-making basis for status maintenance of energy meters and purchasing of spare parts.
出处
《电网与清洁能源》
北大核心
2016年第7期77-80,86,共5页
Power System and Clean Energy
基金
国家电网公司科技项目"支撑互动用电服务的用电信息采集系统技术研究及应用"(524608150061)~~
关键词
非健康值
线性回归
数据挖掘分析
整体状态分析
the non-health value
linear regression
datamining analysis
overall state analysis
