面向电网时序数据的数据质量实时治理技术研究
发布时间：2018-07-28 15:08
【摘要】：电网时序数据是电网设备状态监测、故障诊断的重要基础,对实时性要求较高。然而,现有的数据质量治理方法大多侧重于数据库中已有的历史数据,难以满足实时性的要求,且所采用的方法以及框架由于自身存在的问题,难以应对超大规模的数据集。本文以“面向电网时序数据的数据质量实时治理技术研究”为课题,旨在研究分布式实时计算系统Storm,并将其与时间序列序列分析、数据清洗技术相结合,解决大规模数据集实时治理的问题。本文首先深入研究时间序列分析技术的原理与方法,对时间序列预测模型ARIMA与智能电网时序数据的特点进行分析;其次,对数据质量控制的方法进行归纳总结;最后设计了面向电网时序数据的数据质量实时治理框架以及适用于该框架的时序数据存储模式。本文利用所提出的框架开展针对海量时序数据源的实时并发治理,对时序数据进行预测,比较不同数据样本对预测值的影响,分别采用基于统计与基于聚类的方法,实时识别数据中的孤立点,为电网当前运行状态诊断与未来发展趋势挖掘提供支撑平台。实例从预测精度、运算速度、占用资源等角度验证了本框架的有效性与实用性。
[Abstract]:Power system timing data is an important basis for power equipment condition monitoring and fault diagnosis, and requires high real-time performance. However, most of the existing data quality governance methods focus on the existing historical data in the database, which is difficult to meet the real-time requirements, and the method and framework can not cope with the large scale data set because of its own problems. In this paper, we focus on the research of data quality real-time governance technology for power grid time series data. The purpose of this paper is to study the distributed real-time computing system, and combine it with time series analysis and data cleaning technology. To solve the problem of real-time governance of large-scale data sets. In this paper, the principle and method of time series analysis are studied, and the characteristics of time series prediction model (ARIMA) and smart grid time series data are analyzed, secondly, the methods of data quality control are summarized. Finally, a real-time data quality governance framework for power system timing data is designed. In this paper, we use the proposed framework to implement real-time concurrency governance for massive time series data sources, predict the time series data, compare the effects of different data samples on the predicted values, and adopt statistical and clustering based methods, respectively. Real-time identification of outliers in data provides a supporting platform for current power grid state diagnosis and future trend mining. The effectiveness and practicability of the framework are verified by examples from the angles of prediction accuracy, operation speed and resource occupation.
【学位授予单位】：华北电力大学(北京)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM507
[Abstract]:Power system timing data is an important basis for power equipment condition monitoring and fault diagnosis, and requires high real-time performance. However, most of the existing data quality governance methods focus on the existing historical data in the database, which is difficult to meet the real-time requirements, and the method and framework can not cope with the large scale data set because of its own problems. In this paper, we focus on the research of data quality real-time governance technology for power grid time series data. The purpose of this paper is to study the distributed real-time computing system, and combine it with time series analysis and data cleaning technology. To solve the problem of real-time governance of large-scale data sets. In this paper, the principle and method of time series analysis are studied, and the characteristics of time series prediction model (ARIMA) and smart grid time series data are analyzed, secondly, the methods of data quality control are summarized. Finally, a real-time data quality governance framework for power system timing data is designed. In this paper, we use the proposed framework to implement real-time concurrency governance for massive time series data sources, predict the time series data, compare the effects of different data samples on the predicted values, and adopt statistical and clustering based methods, respectively. Real-time identification of outliers in data provides a supporting platform for current power grid state diagnosis and future trend mining. The effectiveness and practicability of the framework are verified by examples from the angles of prediction accuracy, operation speed and resource occupation.
【学位授予单位】：华北电力大学(北京)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM507
【参考文献】
相关期刊论文 前10条
1 翟静;曹俊;;基于时间序列ARIMA与BP神经网络的组合预测模型[J];统计与决策;2016年04期
2 王远;陶烨;袁军;何卫;;一种基于HBase的智能电网时序大数据处理方法[J];系统仿真学报;2016年03期
3 王远;陶烨;蒋英明;陈波;陈立宇;;智能电网时序大数据实时处理系统[J];计算机应用;2015年S2期
4 韩福霞;储志高;舒彬;刘宏志;尹璐;丁仁山;;基于storm云平台的电力信息系统实时监理的研究[J];电气应用;2015年S1期
5 严英杰;盛戈v，
本文编号：2150639
本文链接：https://www.wllwen.com/kejilunwen/dianlidianqilunwen/2150639.html
