| 摘要: |
| 提出了一种直接面向IEEE COMTRADE格式的海量故障录波数据并行压缩／解压算法。算法给出了COMTRADE数据文件中时间信息无损恢复公式，提出了针对状态量数据的优化RLE编码。对高频模拟量数据采用提升格式小波变换，用硬阈值量化后采用基于LZ77和Huffman混合熵编码的Deflate算法进行压缩，而对低频模拟量数据采用RLE编码。为有效分批压缩COMTRADE数据文件，提出了一种基于高频模拟量数据最佳小波分解层数下的最少补零的数据划分算法。映射到不同CPU核心的压缩／解压线程通过竞争通道序号获取计算数据。简述了压缩文件格式及其采用的变长存储和访问技术。试验证明此算法可获得大压缩比，并且可随着CPU核心数的增加和COMTRADE文件的增加，获得线性加速比。 |
| 关键词: 数据录波 IEEE COMTRADE 压缩 解压 数据处理 LZ77 Huffman Deflate 小波变换 |
| DOI： |
| 分类号: |
| 基金项目:中央高校基本科研业务费专项资金资助项目（ZYGX-2009J089） |
|
| Parallel compression and decompression algorithm for massive recording data in IEEE COMTRADE format |
|
GUI Xun
|
|
School of Aeronautics and Astronautics，University of Electronic Science and
Technology of China，Chengdu 610054，China
|
| Abstract: |
| A parallel compression／decompression algorithm is put forward for the massive fault recording data in IEEE COMTRADE format. A lossless data recovery formula is given for the time information in COMTRADE data file and an optimized coding algorithm is proposed for the status data. The high-frequency analog data are transformed with the lifting wavelet，quantified with hard-threshold，and then compressed by the Deflate algorithm based on the LZ77-Huffman hybrid entropy coding while the low-frequency analog data by the RLE coding algorithm. A data partitioning algorithm is proposed to effectively compress the COMTRADE data file in turn，which is based on the optimal layers of wavelet decomposition for high-frequency analog data for patching the least zeros. Mapped to different CPU cores，the compression／decompression threads access the computing data through the competition of channel serial number. The format of compressed file and the technology of variable-length storage and access are introduced. Test indicates that，the great compression ratio is gained and the linear acceleration ratio is obtained along with the increase of CPU core quantity and COMTRADE file size. |
| Key words: data recording IEEE COMTRADE compression decompression compression decompression data processing LZ77 Huffman Deflate wavelet transforms |
