电力设备载流故障预警与预测的若干关键技术研究

发布时间：2018-06-18 12:31

本文选题：电力设备 + 载流故障　；参考：《浙江大学》2014年博士论文

【摘要】：随着电力系统向大容量、超高压、远距离的方向飞速发展,电力设备的安全可靠运行显得越来越重要。因此,有关电力设备故障诊断技术的研究与应用具有重要的现实意义。怎样提高故障诊断系统的准确性、快速性,使得系统能够快速有效地发现故障、定位故障、分析故障,是电网保护的重要课题。本文针对电力设备的载流故障,在总结现有的故障诊断技术的特点及应用情况的基础上,致力于故障预警及预测,旨在故障发展初期能够发现故障和预测故障的发展趋势。从机理分析、系统建模、数据挖掘等几个方面进行研究,提出了一套电力设备载流故障在线诊断算法,实现了“温度采集-信号传输-数据分析-故障预警-故障预测”一整套流程。与现有诊断系统相比,本文提高了报警准确率、提前报警时间,增加故障预测功能,使得系统在可靠性、快速性及灵敏性上有了显著的提高。基于某真实110kV变电站的应用结果表明了所提方案的有效性。论文的主要内容包括： (1)故障的早期预警。对主成分分析法(PCA)、K-means聚类算法和小波阈值去噪方法进行了简单的介绍。针对载流故障形态多样性的特点,采用上述方法对实时数据和历史数据进行综合分析,研究提出故障特征提取算法和故障识别方法；并通过对邻近位置的相关比较分析,研究提出载流故障的在线早期预警方法。基于PCA的预警方法的实验结果表明,此方法能够准确快速地预报设备故障,并将报警时间大大的提前,具有较高的可靠性和灵敏性。 (2)故障的机理建模。研究分析了载流故障的发展机理,从故障触点发热的根本原因出发,掌握其温度变化的规律。基于传热学原理建立了故障触点的温度模型,并利用最小二乘求得温度模型参数。随后基于温度模型,实现了触点温度拟合、插值以及预测的功能。利用真实温度数据验证所建立的温度模型,结果表明温度模型能够较准确地描述触点温度变化,与实际情况吻合。 (3)故障的趋势预测。对贝叶斯滤波、蒙特卡罗方法、粒子滤波方法进行了详细的介绍,提出了基于粒子滤波的触点温度预测方法。该方法将求和运算代替积分运算,修正温度模型参数,利用修正后的参数代入温度模型,从而进行触点温度预测。每经过一次采样周期,基于更新的温度做一次参数修正,从而得到一条动态变化的预测曲线。应用结果表明,基于粒子滤波方法的预测方法能够准确地揭示触点温度未来的发展趋势,并且随着修正次数的增加,预测曲线与实际曲线逐渐趋向于一致。 (4)介绍了某真实110kV变电站的基本构成,并详细阐述了本文所提方法应用到此变电站的流程。从温度数据的获取、温度数据的分类、温度数据的分析、故障预警及故障预测,应用效果证明了本文所提方法的有效性。
[Abstract]:With the rapid development of power system in the direction of large capacity , ultra high voltage and long distance , the safe and reliable operation of power equipment is more and more important . Therefore , the research and application of fault diagnosis technology of power equipment is of great practical significance . How to improve the accuracy and speediness of the fault diagnosis system can make the system quickly and effectively detect faults , locate faults and analyze faults , which is an important subject of power grid protection .

On the basis of summarizing the characteristics and application of the existing fault diagnosis technology , this paper is devoted to the fault early warning and prediction based on summarizing the characteristics and application of the existing fault diagnosis technology .

The main contents of the thesis include :

( 1 ) In the early warning of fault , a simple introduction to PCA , K - means clustering algorithm and wavelet threshold denoising method are introduced . In view of the characteristics of the diversity of current - carrying faults , the real - time data and historical data are comprehensively analyzed by the above - mentioned method , and fault feature extraction algorithm and fault identification method are proposed .
On the basis of PCA , the results show that this method can predict the fault of equipment accurately and quickly and advance the alarm time greatly , and has higher reliability and sensitivity .

( 2 ) The mechanism of fault is modeled . The mechanism of current fault is analyzed . The temperature model of the fault contact is established based on the principle of heat transfer . The temperature model of the fault contact is established based on the principle of heat transfer , and the function of fitting , interpolation and prediction of the contact temperature is realized based on the temperature model . The temperature model is verified by using the real temperature data . The results show that the temperature model can accurately describe the temperature change of the contact points and agree with the actual situation .

( 3 ) The trend prediction of the fault is introduced . The method of Bayesian filtering , Monte Carlo method and particle filtering is introduced in detail . The method is used to replace the integral operation , the temperature model parameter is corrected , and the temperature model is modified by using the modified parameter , so as to obtain a dynamic dynamic prediction curve . The application results show that the prediction method based on the particle filter method can accurately reveal the development trend of the future of the contact temperature , and the prediction curve and the actual curve tend to be consistent with the increase of the correction times .

( 4 ) The basic structure of a real 110kV transformer substation is introduced , and the process of applying the method to this substation is described in detail . From temperature data acquisition , temperature data classification , temperature data analysis , fault early warning and fault prediction , the effectiveness of the proposed method is proved .
【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TM507

【参考文献】