电力设备载流故障趋势预测技术研究

发布时间：2018-07-31 19:43

【摘要】：电力设备的可靠运行关乎超大规模输配电和国家电网的安全,因此,有关电力设备可靠性的研究具有重要的现实意义。电力设备的故障诊断是保证设备安全运行的重要举措,因而成为电力系统和自动化领域的重要研究课题。电力设备故障是危及电网安全运行的重要因素,而载流故障又是电力设备故障的主要类型之一。本文针对电力设备的载流故障,从机理分析、数据挖掘等几个方面着手,利用神经网络、混沌时间序列分析等技术成果,提出了一套完整的电力设备载流故障诊断算法。对来自某电厂的实际运行温度数据进行分析,电力设备载流故障从形态上主要可以分成两类,温度快速上升并超限的故障形态和温度缓慢振荡上升并超限的故障形态。针对不同的故障形态,研究提出相应的载流故障诊断算法。对于温度快速上升并超限的故障形态,本文提出了一种基于主元分析法(PCA)和K-means聚类算法的载流故障早期预警算法。然后,利用极限学习机(ELM)预测故障触点的温度序列,极限学习机(ELM)具有学习速度快、泛化性能好的优点。应用在本系统当中,能够在保证一定预测精度的前提下,更好地满足系统的实时性要求。对于温度缓慢振荡上升并超限的故障形态,本文提出了基于多尺度混沌时间序列预测的载流故障趋势预测方法。首先应用基于小波变换的噪声平滑算法对原始数据进行降噪处理,并利用降噪后的数据构造即时温度序列和多时间尺度的平均温度序列。通过将变步长的归一化最小均方误差算法应用于Volterra核函数的辨识中,提高核函数的辨识精度,从而获得更加准确的预测效果；采用上述改进后的Volterra自适应滤波算法对载流故障进行趋势预测。基于某电站实际运行数据的测试结果表明了所提方法的有效性。
[Abstract]:The reliable operation of power equipment is related to the safety of super large scale transmission and distribution and the state grid. Therefore, the research on the reliability of power equipment has important practical significance. Fault diagnosis of power equipment is an important measure to ensure the safe operation of power equipment, so it has become an important research topic in the field of power system and automation. Power equipment fault is an important factor that endangers the safe operation of power network, and the current carrying fault is one of the main types of power equipment fault. Aiming at the current carrying fault of power equipment, this paper presents a complete diagnosis algorithm for current carrying fault of power equipment from the aspects of mechanism analysis, data mining and other technical achievements, such as neural network, chaotic time series analysis and so on. Based on the analysis of the actual operating temperature data from a power plant, the current carrying faults of power equipment can be divided into two types: the fault pattern of temperature rising rapidly and exceeding the limit, and the fault pattern of temperature oscillating slowly rising and exceeding the limit. According to different fault patterns, the corresponding current-carrying fault diagnosis algorithm is proposed. In this paper, an early warning algorithm of current-carrying faults based on principal component analysis (PCA) and K-means clustering algorithm is proposed for the fault pattern of temperature rising rapidly and exceeding the limit. Then, the extreme learning machine (ELM) is used to predict the temperature sequence of the fault contact. (ELM) has the advantages of fast learning speed and good generalization performance. Application in this system can better meet the real-time requirements of the system on the premise of certain prediction accuracy. In this paper, a method for predicting the trend of current-carrying faults based on multi-scale chaotic time series is proposed for the fault pattern of temperature rising slowly and exceeding the limit. Firstly, the noise smoothing algorithm based on wavelet transform is used to reduce the noise of the original data, and the real time temperature series and the average temperature series with multiple time scales are constructed by using the de-noised data. By applying the normalized minimum mean square error algorithm with variable step size to the identification of Volterra kernel function, the accuracy of kernel function identification is improved, and a more accurate prediction result is obtained. The improved Volterra adaptive filtering algorithm is used to predict the trend of current carrying fault. The test results based on the actual operation data of a power station show the effectiveness of the proposed method.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TM507

【相似文献】