基于压缩感知理论的风电变流器检测信号处理方法研究

发布时间：2018-04-24 12:01

本文选题：变流器 + 压缩感知　；参考：《兰州交通大学》2017年硕士论文

【摘要】：目前,风电变流器检测信号处理方法均建立在奈奎斯特采样定理的基础上,这将产生巨大的采样数据,使得数据的存储或者传输引起“空间灾难性”后果。压缩感知(compressed sensing,CS)理论以远低于奈奎斯特采样频率采集信号,且信号的采样和压缩同时进行,以信号的非自适应投影来保留原始信号的关键信息,通过数学方法来准确重构原始信号。针对以上研究背景,本文紧紧围绕压缩感知理论对风电变流器检测信号进行处理,该方法既具有压缩感知的压缩采样特性,又具有准确恢复重构的能力,因此具有重要的研究价值。首先,研究了压缩感知理论,搭建了变流器仿真模型,分析了变流器原始电压检测信号的稀疏性。由稀疏性信息,研究电压信号稀疏变换基与投影矩阵之间的关系,由优化投影理论,降低互相关性来优化投影矩阵,该方法使得投影矩阵的压缩性能提升。其次,为了解决直接利用CS理论对变流器输出端三相电压检测数据存储空间浪费以及重构性能差等问题,利用三相电压的关系,研究了基于坐标变换的三相电压检测信号CS处理方法。先对输出端三相电压检测数据采用坐标变换后转化为一维信号,再利用CS理论对其进行压缩重构,重构的信号将其转化为两相信号并作坐标反变换,即得到重构的三相电压信号。实验表明,该方法处理变流器电压检测信号时,可有效压缩原始三相电压数据,使得运行时间更低、重构误差更小,且节约了测量数据的存储空间。然后,对比研究了五种典型的贪婪重构算法。针对原始信号重构性能差的问题,研究了基于广义Jaccard系数的广义正交匹配追踪算法,即就是利用广义Jaccard系数相似性准则去替换内积相似性准则来对支撑集进行优化。在相同条件下与其它的重构算法相比较,改进的算法所具有的重构性能更优。最后,考虑到电压检测信号所蕴含的深层信息等问题,研究了基于本征时间尺度分解(intrinsic time-scale decomposition,ITD)与改进内积的压缩感知重构算法相结合的风电变流器电压信号重建方法。基于ITD把变流器电压检测信号处理成互不影响的合理旋转分量和余量,进而利用基于广义Jaccard系数的广义正交匹配追踪算法对每个分量进行处理,重构合并得到变流器原始电压检测信号。该方法降低了计算复杂度和重构误差。
[Abstract]:At present, the detection signal processing methods of wind power converter are based on Nyquist sampling theorem, which will produce huge sampling data, which makes the storage or transmission of data cause "space catastrophic" consequences. Compressed sensing theory collects signals at a much lower sampling frequency than Nyquist, and the sampling and compression of signals are performed simultaneously. The key information of the original signal is preserved by the non-adaptive projection of the signal. The original signal is reconstructed by mathematical method. In view of the above research background, this paper focuses on the compression sensing theory to process the detection signal of wind power converter. This method not only has the compression sensing characteristic of compression sampling, but also has the ability of accurate recovery and reconstruction. Therefore, it has important research value. Firstly, the compression sensing theory is studied, the converter simulation model is built, and the sparsity of the original voltage detection signal is analyzed. Based on the sparse information, the relationship between the sparse transformation basis of voltage signal and the projection matrix is studied. The projection matrix is optimized by optimizing the projection theory and reducing the mutual correlation. The compression performance of the projection matrix is improved by this method. Secondly, in order to solve the problem of waste of data storage space and poor reconfiguration performance of three-phase voltage detection of converter output by using CS theory directly, the relationship of three-phase voltage is used. The CS processing method of three-phase voltage detection signal based on coordinate transformation is studied. The three-phase voltage detection data at the output end are transformed into one-dimensional signal by coordinate transformation, then compressed and reconstructed by CS theory. The reconstructed signal transforms the signal into two-phase signal and makes coordinate inverse transformation. The reconstructed three-phase voltage signal is obtained. The experimental results show that this method can effectively compress the original three-phase voltage data, make the running time lower, the reconstruction error smaller, and save the storage space of the measurement data. Then, five typical greedy reconstruction algorithms are compared. Aiming at the poor performance of the original signal reconstruction, the generalized orthogonal matching tracking algorithm based on generalized Jaccard coefficients is studied, that is, the generalized Jaccard coefficient similarity criterion is used to replace the inner product similarity criterion to optimize the support set. Compared with other reconstruction algorithms under the same conditions, the improved algorithm has better reconfiguration performance. Finally, considering the deep information contained in the voltage detection signal, the voltage signal reconstruction method of wind power converter based on intrinsic time scale decomposition intrinsics time-scale decompositionITD and improved internal product compression sensing reconstruction algorithm is studied. Based on ITD, the voltage detection signal of converter is processed into reasonable rotation component and residual, and then each component is processed by generalized orthogonal matching tracing algorithm based on generalized Jaccard coefficient. The original voltage detection signal of the converter is obtained by reconstructing and merging. This method reduces computational complexity and reconstruction error.
【学位授予单位】：兰州交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM614;TM46

【参考文献】