近年来迅速发展起来的小波变换分析方法，具有多尺度分析和时－频局部化特性，特别适合于电力系统暂态突变信号和微弱变换信号的处理和特征抽取。 由于保护的速动性，提取的信号中除含有工频、丰富的谐波分量、高频外，还有衰减的非周期分量。虽然也采取了一些数字滤波措施，但大多数都是把衰减的非周期分量看作稳态直流或线性化处理，都不可避免地带来误差。就是目前被广泛使用的集成了数字滤波和保护算法的傅氏算法本身还会受到非整次谐波的影响。本文研究了基于小波变换的数字滤波及求取工频信号幅值的保护算法。仿真分析证明了此方法能极大地滤除衰减直流分量、噪音和非整次谐波的影响，并通过与传统的傅氏算法比较，验证了该算法能很好地提高算法的精度，减小误差，增加保护的可靠性和灵敏性。 在变压器微机保护中，关于励磁涌流和短路电流的识别是困扰继电保护研究人员的棘手问题。文中对励磁涌流产生的机理进行了分析，并对励磁涌流和短路电流进行了数字仿真，比较了励磁涌流和短路电流在物理特性上的区别。随后，根据小波变换后励磁涌流在高频率段较内部故障呈现出明显的奇异性，本文研究了基于一尺度下的小波系数奇异度和值判据，用来提取故障特征，识别励磁涌流的。大量的EMTP仿真数据验证了判据的正确性、可行性，并能提高微机保护装置在轻微故障时启动的灵敏性，解决了三相二次谐波综合制动的保护原理在变压器空合于内部故障时延时动作甚至拒动的难题

In recent years, the wavelet transform analysis method has been developed rapidly. As the wavelet transform possesses the properties of multi-scale analysis and time-frequency domain localization, it is particularly well adapted to transient sharply changing signals and slight changing signals, and to extracted characteristics of these signals. Owing to relay protection acts promptly, extracted signals are including abundant harmonious, high-frequency and damping non-periodic components. Although some digital filtering measures are also adopted, it is an inevitable error because the damping non-periodic components is considered as steady direct current or line component to deal with on most occasions. And that, at present used wildly Fourier algorithm is affected by non-integral harmonious that integrates digital filtering and protection algorithm. This paper studies digital filtering and protection algorithm for peek value of base frequency based on Wavelet Transform. Compared with the traditional Fourier algorithm, it is more reliable and sensitive due to it can consumedly filter damping non-periodic component, noises, and non-integral harmonious. It is very important and difficult for investigator to identify magnetized inrush currents and internal faults for large transformers relay protection. Firstly, this thesis analysis the mechanism of magnetizing inrush currents in detail, and get the difference in physical characteristic between inrush currents and internal faults. Subsequently, by studying the waveform features extracted by wavelet transform, this thesis proposes a method that can discriminate inrush currents from internal faults. The extensive EMTP simulation results have verified the method is a reliable and feasible way, and can improve sensitivity of computerized protection devices when slight faults happen, also have drastically solved a problem that protection devices act in delay even in error as an internal fault transformer is closed emptily.