基于高维数学形态学和形态学小波的电力系统故障信号特征提取

Power system fault signal feature extraction is an important issue for protective relaying and power quality analysis. However, for over half a century it has been relied on traditional signal processing techniques such as Fourier transform and Wavelet transform. They focus on the frequency information of a signal by transferring it to the frequency domain, which requires complex integral calculation, and cannot reflect the shape information of the signal or obtain the knowledge of the transients. Mathematical morphology (MM), on the other hand, processes and analyses the geometrical structures of a signal, yet it is mainly concerned with image processing. Its potential has not been fully explored for applications in a high dimensional domain, and its ability for frequency analysis has never been attempted. In order to tackle these issues, this project will propose fault signal feature extraction algorithms based on high-dimensional MM and morphological wavelet, and will research on the topics of space transform, high-dimensional morphological operators, feature extraction, morphological wavelet, and MM's relationship between time and frequency domains. Targeting the characteristic of power system fault signal, a set of transient feature extraction algorithm will be developed, in order to solve the problems such as DC offset removal, saturation detection and compensation of secondary current, disturbance detection, location and classification, harmonics estimation, etc. The research on high-dimensional MM and MM's ability on frequency analysis will explore a new realm and will have a magnificent impact on the field of signal processing.

电力系统故障信号特征提取是继电保护和电能质量分析的重要问题。然而半个多世纪以来，对其处理一直依赖于傅里叶变换和小波变换等传统方法。这些方法需要通过复杂的积分计算将信号转换到频域进行分析，且无法反映信号的瞬时信息。数学形态学主要分析和处理信号的波形，然而其绝大多数应用为图像处理，在高维空间中的应用尚有待开发，且缺乏对频率信息的分析。针对这些问题，本项目提出基于高维数学形态学和形态学小波的故障信号特征提取方法，将对空间变换、高维数学形态学、特征提取、形态学小波、形态学在时频域间沟通等问题进行深入理论研究，针对电力系统故障信号的特点开发快速准确的瞬时特征提取算法，拟解决的问题包括去除故障信号的直流偏移、检测和补偿二次电流饱和、检测暂态信号、检测和定位电能扰动并进行分类、分析谐波成分。另外对高维数学形态学及数学形态学在时频域间的联系的研究是一个崭新的方向，研究成果会对信号处理领域产生重要影响。

电力系统故障信号特征提取是继电保护和电能质量分析的重要问题。然而半个多世纪以来，对其处理一直依赖于傅里叶变换和小波变换等传统方法。这些方法需要通过复杂的积分计算将信号转换到频域进行分析，且无法反映信号的瞬时信息。数学形态学主要分析和处理信号的波形，然而其绝大多数应用为图像处理，在高维空间中的应用尚有待开发，且缺乏对频率信息的分析。针对这些问题，本项目提出基于高维数学形态学和形态学小波的故障信号特征提取方法，对空间变换、高维数学形态学、特征提取、形态学梯度、形态学小波、高级形态学算子等问题进行深入理论研究，针对电力系统故障信号的特点开发快速准确的瞬时特征提取算法，完成了电流互感器二次电流饱和的检测和补偿、励磁涌流及和应涌流的检测和识别、谐波估计、直流偏移去除、电能质量扰动信号的检测与分类、风能预测等工作，所开发的算法分别在MATLAB、PSCAD、RTDS仿真信号及实际信号上进行了验证，实验结果证明了所开发的算法的准确性和有效性。本项目的主要研究成果已形成论文和专利，共发表和录用期刊论文7篇（其中SCI索引6篇、EI索引1篇），国际会议论文13篇（均为EI索引），申请发明专利13项（其中5项已授权）。