【作者】 桂勋；

【导师】 钱清泉；

【作者基本信息】 西南交通大学 ， 电力系统及其自动化， 2009， 博士

【摘要】 随着电力系统自动化的快速发展和COMTRADE数据的日益广泛应用,基于COMTRADE的高级电力系统故障分析软件在电力系统安全运行中变得越来越重要。然而,随着电子技术的快速发展,电力系统暂态录波逐步向高采样率、连续记录和海量存储的方向发展,COMTRADE数据的海量化趋势明显,并已经在电力系统应用中造成了各种亟不可待的问题。同时,目前多核技术已经大规模普及,引发了并行计算的技术革命,这就为解决由海量COMTRADE数据引发的各种问题,提供了解决思路。另外,在软件设计方法学方面,模式理论已经成为当今指导先进软件系统开发的重要方法学。于是,本论文将多核并行计算技术和模式方法学相结合,首次研究并提出了针对海量COMTRADE数据处理和计算的一系列并行算法和快速算法,并采用当今先进的计算机技术,构建了一整套适应电力系统未来发展的高级电力系统暂态信息并行处理和分析系统(一套全图形化故障分析软件,一套可跨平台的在线自动故障分析软件,一套可跨平台的COMTRADE并行压缩/解压软件)。本文的研究成果解决了该领域内各种棘手的关键算法和技术问题,为电力系统海量暂态信息处理开辟了新的途径。目前,该系统已投入实际使用。在海量COMTRADE数据处理方面,通过深入研究COMTRADE标准和大量现场数据,提出了COMTRADE容错解析算法,大量试验证明该算法可明显提高故障分析软件的鲁棒性。为解决海量COMTRADE数据加载过程中效率低下的问题,采用并行流水线和共享队列的设计思想,并利用以往串行加载算法中没有用到的序号信息,针对二进制和ASCⅡ码数据文件分别提出了各自的并行加载算法,试验证明该算法可显著提高海量COMTRADE数据的加载效率。为解决海量COMTRADE数据大量占据磁盘空间和传输带宽的问题,采用多种提升格式小波和熵编码算法以及线程间竞争通道序号的思路,提出了面向COMTRADE格式的海量录波数据并行压缩/解压算法,试验证明该算法可获得较大的压缩比,且速度极快。为解决海量COMTRADE波形数据绘制效率低下的问题,以分图层并行绘制而后融合的思路,提出了海量COMTRADE波形数据并行绘制算法,试验证明了此算法可显著提高海量COMTRADE波形数据的绘制效率。在海量COMTRADE数据计算分析方面,研究了快速傅里叶算法的并行化和加窗双谱线插值谐波分析算法,试验证明了该算法的有效性。在分析相关文献的基础上,提炼出了适合海量COMTRADE通道数据计算的各种递推算法,并进一步提出了其通用的并行计算算法,试验证明了该算法可显著提高海量COMTRADE通道数据的计算效率。通过研究高压输电线路线电压幅值沿线分布规律,得出了现实中该分布曲线只可能最多由两条单调方向不同的曲线段组成的结论,并利用此结论提出了一种可排除伪根、精确求解、大幅降低计算量和简化程序设计的故障双端测距算法,试验证明了该算法的有效性。在软件系统研究方面,为解决目前各种第三方电力系统故障分析软件中存在的一系列不足和缺陷,采用面向对象的设计模式方法学,提炼出了适合电力系统故障分析软件设计的各种模式。为提高全图形分析环境的分析效率,结合设计模式,提出了一种和海量通道数据无关的图形记忆算法,实现了Undo/Redo功能。为提高电力系统故障分析的效率和直观性,提出了其三维分析方法,给出了故障分析中三维矢量数据的生成算法。为解决现有故障分析软件中二次分析环境的不足和缺陷,研究并设计了可扩展的电力系统故障集成分析环境。环境内集成了大量的电力系统通用分析算法,并可采用插件技术进行算法扩展。为和保护信息系统进行无缝融合,实现在线自动故障分析,研究并设计了一套可跨平台的COMTRADE在线自动故障分析软件系统,并提炼出与其相关的各种跨平台关键技术和算法。在软件体系结构研究方面,为指导未来更加先进电力系统专业分析软件的开发,结合动态语言技术和面向服务的体系结构(SOA),提出了一种可适应未来发展、具有良好扩展性的电力系统通用分析软件体系结构。更多还原

【Abstract】 With the rapid development of power system automation and increasingly wider application of COMTRADE data of power system, the advanced analysis software system of power system based on COMTRADE is playing more important role in the safe operation of power system. However, combined with the rapid development of electronic technical, transient data record of power system is gradually growing toward higher sample rate, continuous record and massive storage, the massive trend of COMTRADE data is extremely obvious so that many exigent and important problems in the application of power system are resulted in. Meanwhile, as the multi-core technical is becoming more popular and causes a technical revolution of parallel computation, it provides feasible solution for the application problems caused by massive COMTRADE data. In addition, in the aspect of software design methodology, pattern theory becomes important one to guide the implementation of advanced software system. In this thesis, through the combination of multi-core technical and pattern methodology, a series of parallel algorithms and fast algorithms for the processing and computation of massive COMTRADE data are firstly put forward; by usage of these algorithms and current advanced computer technical, a set of advanced parallel processing and analysis systems of power system transient information are implemented (including a all-graphic fault analysis software system, a cross-platform online automatic fault analysis software system and a cross-platform COMTRADE parallel compression and decompression software system). The main contributions of this work have solved the stubborn and key algorithms and technical problems in power system. Now, these systems are put into service in power system.In the aspect of massive COMTRADE data processing, through the deep research on COMTRADE standards and a lot of field files, a fault-tolerance parsing algorithm of COMTRADE is proposed. Plentiful testing results of field data indicated that the new algorithm could obviously improve the robustness of fault-analysis software. To solve the problem of low efficiency in course of loading massive COMTRADE data file, by usage of the design idea of parallel pipeline and share queues, as well as the sequence number omitted in the past series-loading algorithm, two individual parallel loading algorithms are proposed for data files of BINARY and ASCII format. The testing results indicated that the algorithms could greatly improve the loading efficiency of massive COMTRADE data. To solve the problems of large disk space and transmission bandwidth occupied by massive COMTRADE data, by the adoption of various lifting scheme wavelet transform and entropy coding algorithms, as well as the idea of competition channel serial number among threads, a parallel compression and decompression algorithm of massive COMTRADE recorded data is proposed. The testing results indicated that great compression ratio could be got for massive COMTRADE file and its compression speed was extremely fast. To solve the problem of low efficiency to paint massive COMTRADE waveform data, by usage of the idea of parallel paint by graphic layer then mergence, a parallel painting algorithm for massive wave COMTRADE waveform data is provided. The testing result indicated that it could greatly improve the efficiency to paint the massive COMTRADE waveform data.In the computation and analysis aspect of massive COMTRADE data, the parallel fast Fourier transform algorithm and harmonic analysis algorithm based on window double-spectrum-line interpolation are researched. The testing result indicated it was feasible. On basis of analysis of a lot of related references, various recursive algorithms suitable for the computation of massive COMTRADE channel data is provided, furthermore, its general parallel computation algorithm is also proposed. The testing result indicated that it could greatly improve the computation efficiency of the massive COMTRADE channel data. By analyzing the change of line voltage along HV transmission line, conclusion was drew that the line voltages curve along HV transmission line only consisted of two different monotonous direction curve segments in the actual length line at most. With this conclusion, two-terminal fault location algorithm is proposed. It can remove false root andsolve accurately,as well as greatly reduce the computation amount and simplify the program design.In the research aspect of software system, to solve the defects and shortcomings existed in fault analysis software systems provided by third party, by usage of object-oriented design pattern methodology, various design patterns suitable for fault analysis software in power system are summarized. To improve the analysis efficiency in all-graphic analyzing environment, by combination of design pattern methodology, a graph state memory algorithm that is independent of massive channel data is proposed. The Undo/Redo function is realized. To improve the efficiency and intuitiveness, a 3-D method for fault analysis in electric power system is proposed and the generation algorithm of 3-D vector data for fault analysis is given. To solve the defects and shortcomings of secondary analysis environment in current fault analysis software systems, expansive integrated analysis environment for power system fault is researched and designed. Amount of general power analysis algorithms are integrated in this environment and plug in technical is used to expand the analysis algorithm. To integrate seamlessly with protection information system and realize the fault analysis automatically, a cross-platform online automatic fault analysis software system is researched and designed; some related important cross-platform technologies are summarized.In the research aspect of software structure, to guide the research and development of more advanced professional analysis software in future, by combination of dynamic language technical and Service Oriented Architecture (SOA), a general architecture of power system analysis software is proposed. It is suitable for future development and has remarkable extensible characteristic.更多还原