摘要
为提高局部放电信号的消噪效果,根据统计学习理论中的结构风险最小化思想,基于小波的多分辨分析技术,给出一种针对局部放电的消噪方法。该方法不同于以往的对噪声进行估计并加以阈值处理的消噪方法,而从估计真实放电信号的角度出发,根据小波分解的系数能量大小选择能够表征真实局放信号的小波系数,并通过结构风险最小化原则确定模型的VC维,从而获得最优的真实信号估计。仿真结果表明,该方法在信噪比和均方误差指标上均要高于传统的史坦无偏似然估计阈值方法(SURE)、固定阈值法(VISU)及最小最大阈值方法(Min-iMax)。实验室实验和现场实测数据的处理结果表明,该方法能够还原真实的放电信号,提高检测信号的信噪比,是一种对局部放电信号消噪的较优算法。
To suppress the interferences in the partial discharge signals,a new signal denoising method for partial discharge based on structural risk minimization(SRM) principle is introduced.Unlike traditional threshold method by estimating the noise level,the new method tries to suppress the interference by giving an accurate estimation of the real discharge signal.Based on the wavelet decomposition of the signal,decomposition coefficients are selected according to their descending ordered energy.Under the inductive principle of SRM,the VC dimension of the denosing structure,which is also the number of wavelet coefficients alternatively,can be determined.Four partial discharge pulses with white noise of four different intensities are thus simulated.The simulation results show that the mentioned method gives a better performance than SURE,VISU and MiniMax wavelet threshold in the comparison of mean square error and signal-noise ratio.The proceeding results of the signals collected from lab test and field test also show that the mentioned method improves the signal-noise ratio significantly.
出处
《高电压技术》
EI
CAS
CSCD
北大核心
2011年第5期1172-1179,共8页
High Voltage Engineering
关键词
局部放电
结构风险
统计学习
VC维
小波变换
分解系数
partial discharge
structural risk
statistical learning
VC dimension
wavelet transform
decomposition coefficient