中文摘要：

电能质量信号去噪是电能质量扰动判决和定位的基础。电力系统中，绝大部分电能质量信号都含有较明显的突变点。该文首先分析了现有各种去噪算法的不足，然后在此基础上针对信号的二阶导数设计了。个局部滑动窗，并将该滑动窗分成左右两个长度相等的窗口，最后以左右窗口均值差的似然比作为判决准则来判决突变点。基于突变点判决的结果，对突变点区域和非突变点区域采用不同的滤波方法。实验结果表明，该文所提出的混合滤波算法，去噪效果良好，能较完整的保留突变点的信息，较好地解决了保护信号局部特征与抑制噪声之间的矛盾，在绝人部分电能质量去噪领域，有突出的优势。

英文摘要：

De-nosing of power quality signals is the base of detecting and locating power quality disturbances. In the electric power system, most power quality signals contain obvious break points. Firstly shortcomings of various kinds of present de-noising methods are analyzed. Then a local sliding window is designed for second derivative signals, what＇more, it is divided into two windows and the length of the left window is equal to the right one. Finally, the likelihood ratio of mean difference between the left window and the right one is taken as a decision criterion to detect break points and a novel filter based on it is designed. Two usual signals-voltage interruptions and voltage notches are processed with this method, and the results are compared with the ones of the average filter and the median filter. Experiment results show that the de-noising effect of this method is good and it preferably solves the contradiction between protecting local character of signals and restraining noises. In most de-nosing fields of power quality signals, this method has prominent advantage.