| 引用本文: | 杨鹏,申洪涛,陶鹏,冯波,张洋瑞,王立斌.云平台下时间序列数据并行化排列熵特征提取方法[J].电力自动化设备,2019,39(4): |
| YANG Peng,SHEN Hongtao,TAO Peng,FENG Bo,ZHANG Yangrui,WANG Libin.Parallel permutation entropy feature extraction method for time series data based on cloud platform[J].Electric Power Automation Equipment,2019,39(4): |
|
| 摘要: |
| 随着高级量测体系和各类监控系统的大规模建设发展，时间序列数据的规模呈指数级增长，在智能电网大数据中占有较大的比重。时间序列数据的特征提取是影响数据挖掘质量的关键步骤，在大数据背景下，传统的特征提取算法已无法满足海量数据处理的需求。结合云计算平台和MaxCompute大数据处理技术，设计实现了时间序列数据的表存储方法和并行化的时间序列数据排列熵特征提取算法。在云计算平台上采用不同规模的数据集对并行化算法进行测试，验证了并行化排列熵算法的正确性和高性能。 |
| 关键词: 时间序列数据 排列熵 特征提取 并行算法 大数据 云计算 |
| DOI：10.16081/j.issn.1006-6047.2019.04.032 |
| 分类号:TM732 |
| 基金项目: |
|
| Parallel permutation entropy feature extraction method for time series data based on cloud platform |
|
YANG Peng1,2, SHEN Hongtao1,2, TAO Peng2, FENG Bo2, ZHANG Yangrui2, WANG Libin2
|
|
1.State Grid Hebei Energy Technology Service Limited Company, Shijiazhuang 050000, China;2.State Grid Hebei Electric Power Research Institute, Shijiazhuang 050000, China
|
| Abstract: |
| With the large-scale construction and development of AMI(Advanced Metering Infrastructure) and various monitoring systems, the size of time series data grows exponentially, which occupies a large proportion in the smart grid big data. The feature extraction of time series data is a key step that affects the quality of data mining. Traditional feature extraction algorithms can no longer meet the requirements of mass data processing in the context of big data. The table storage method and feature extraction algorithm based on parallel permutation entropy are designed and implemented for time series data by combining with the cloud computing platform and MaxCompute big data processing technology. Different scale data sets are tested on the cloud computing platform, and results verify the accuracy and high performance of the parallel permutation entropy algorithm. |
| Key words: time series data permutation entropy feature extraction parallel algorithm big data cloud computing |
