【作者】 熊卫华；

【导师】 赵光宙；

【作者基本信息】 浙江大学 ， 控制理论与控制工程， 2006， 博士

【摘要】 随着电力系统自动化程度的提高,对故障信号处理的精确性和准确性提出了新的要求。传统时频分析方法在处理复杂的故障信号时存在一定的局限性,已经不能满足电气设备在线状态监测和故障诊断技术进一步发展的需求。因此,以现代信号处理技术为基础的设备运行状态监测和故障诊断技术在保障电力系统安全运行,预防事故发生等方面具有重要的意义。 本文深入研究了经验模态分解方法,并围绕经验模态分解方法在电气设备状态评估和故障诊断中的应用展开讨论,主要工作包括: 1 研究了采用经验模态分解方法,筛选信号的本征模态函数,提取信号局部特征的方法。采用该方法提取信号的本征模态函数,对本征模态函数进行希尔伯特变换得到“时间-频率-能量”三维希尔伯特谱。通过和小波变换相比较,说明了基于经验模态分解方法的希尔伯特谱具有更高的分辨率和集中度,更好地反映了信号的时频分布。 2 基于模型延拓非平稳信号端点,提出了处理经验模态分解边界问题的方法。由于仅依据观测区间内部的极值点描述信号上下包络,会给经验模态分解过程带来边界误差,而且边界误差随着分解过程的进行向内传播,进一步污染内部数据,引起分解结果的不合理。采用该方法,获得观测区间外的极值点参与经验模态分解,可以有效地控制边界误差。 3 基于非均匀B样条曲线插值数据极大值、极小值点,提出了经验模态分解信号包络拟合的新方法。该方法通过规范积累弦长参数化,得到定义域内的节点矢量,利用信号极大值、极小值点反算得到非均匀B样条曲线的控制多边形,一起构造非均匀B样条曲线,参与经验模态分解,拟合信号包络。采用该方法可以获得精确的瞬时平均值,从而抑制没有意义的信号波动,避免分解过程中出现因三次样条曲线插值而引起的过冲、欠冲以及不完全包络等问题。更多还原

【Abstract】 At present, the requirement of failure information processing technique is growing gradually with the rapid development of electric power automatization. However the traditional signal processing technique has limitation in accuracy and veracity, truthless characteristic hampers the research on on-line condition monitoring and fault diagnosis for electrical equipment. So condition monitoring and fault diagnosis system based on modern signal processing plays more and more important role in accident prevention and stable operation of power system.In this dissertation, empirical mode decomposition method is studied. Besides, condition evaluation and fault diagnosis for electrical equipment using empirical mode decomposition method is discussed. The major contents are shown as follows:1 A new method of obtaining local characteristic of data based on empirical mode decomposition(EMD) is introduced. EMD separates the time series into a finite and small number of intrinsic mode functions(IMF), which is directly decomposed from time series and could reflect the intrinsic physical property of signal more clearly. Applying the Hilbert transform to IMF, the time-frequency-energy distribution is got. By comparing with wavelet transform, the power of this method is demonstrated.2 A method of forecasting non-stationary signal by model is developed, empirical mode decomposing data is extended using this method. According to nothing but the extrema inside observation data, envelope fitting can be got with error on the borders. The end effects will be serious along with empirical mode decomposition. Data extending method for handling the end effects of EMD is effective.3 A new method of envelope fitting for empirical mode decomposing databased on non-uniform B-spline curve is developed. Firstly, select cumulative chord length parameterization and computing the knots of B-spline. Secondly, determine control vertexes and control polygon. Then, B-spline interpolation curves are obtained. Envelope fitting based on non-uniform B-spline curve handles the incomplete envelope effects of EMD.4 The vibration mechanism of transformer body is discussed. Faults of the winding and core mostly lead to transformer accident, especially delitescent faults would lead to casualty with cumulating result of vibration when transformers are operating. Through research on the relationship between the conditions of winding and core and their vibration, a new method of obtaining the connection between the energy distribution characteristic and the conditions of transformer’s core and winding based on EMD is developed. Associating with forecasting vibration data beyond observation series by model, envelope fitting of data inside and outside observation series based on non-uniform B-spline curve is got. EMD can use this envelope for obtaining intrinsic mode functions. EMD based on forecasting model and B-spline interpolation curves is good for evaluating conditions and diagnosing failures of power transformers.更多还原