220kV油浸电力变压器绕组变形在线诊断方法研究

发布时间：2018-04-16 11:15

本文选题：变压器 + 绕组变形　；参考：《沈阳工业大学》2017年硕士论文

【摘要】：在电力系统中,电力变压器是输变电设备中的核心,连接着多个电压等级,在电网中处于枢纽地位。根据国家电网及南方电网统计数据:2004年国家电网110kV以上电力变压器按部位损坏统计显示,绕组损坏台次为37台,占总数的69.8%;广西电网2013年总损坏台次5次,绕组损坏台次为4台次,占比80%。绕组的损坏多数不是绝缘问题,而是机械性能不足导致的电气故障,对变压器的稳定运行影响巨大。绕组变形具有隐蔽性、累加性,因此实现绕组变形的在线诊断具有重要意义。本文在变压器振动机理研究的基础上,从离线试验数据的分析入手,而后研究在线数据的分离诊断,利用振动法实现了变压器绕组变形的在线诊断。在此研究的基础上,利用基于短路电抗及振动特征量的多信息诊断模型增加了绕组变形诊断的准确性与可信度。本文涉及的试验包括试验变压器空载及负载试验、试验变压器短路冲击试验、220kV电力变压器在线长时间监测试验,获取了大量的研究数据。离线试验的数据分析表明,短路冲击对于铁芯振动的影响较小,经过三轮短路冲击后,其振动信号频域分布变化不大,频域幅值的标准偏差分别为0.0711、0.0778、0.0775。本文提出使用小波包分解后得到的“频段—能量—欧氏距离”作为振动特征值来量化绕组的变形情况。以A相为例,对负载试验得到数据进行分析,三轮冲击试验后其特征值依次为0.0648、0.1094、0.2062。本文通过盲源分离算法对在线变压器振动信号进行信号的分离,通过仿真数据及挂网数据的分析,验证了算法适用于变压器振动信号的分离。通过神经网络训练网络来达到智能识别分离后绕组振动信号的目的。对三台220kV变压器挂网运行历史进行统计,对比分析同负荷下的振动信号,验证了“频段—能量—欧氏距离”作为振动特征量亦适用于220kV变压器,但是其诊断阈值与试验变压器不同。单一数据诊断的结果存在着很多的不足,可信度与准确度不高。在实现变压器绕组变形振动法检测的基础上,本文通过基于短路电抗与振动特征分析的多信息诊断模型诊断变压器绕组变形的状态,数据分析的结果证明其有更高的准确度,对较小变形也有分辨能力。
[Abstract]:In power system, power transformer is the core of power transmission and transformation equipment, connected with many voltage levels, and is in the pivotal position in the power network.According to the statistical data of State Grid and Southern Power Grid: in 2004, according to the statistics of site damage of power transformers above 110kV in State Grid, the number of windings damaged was 37, accounting for 69.8% of the total, and the total number of power transformers damaged in Guangxi Power Grid in 2013 was 5 times.Winding damage stage for 4 times, 80%.Most of the windings damage is not insulation problem, but the electrical failure caused by insufficient mechanical properties, which has a great impact on the stable operation of transformers.Winding deformation has concealment and accumulation, so it is important to realize on-line diagnosis of winding deformation.Based on the research of transformer vibration mechanism, this paper starts with the analysis of off-line test data, then studies the separation diagnosis of on-line data, and realizes the on-line diagnosis of transformer winding deformation by vibration method.On the basis of this research, the accuracy and reliability of winding deformation diagnosis are improved by using multi-information diagnosis model based on short-circuit reactance and vibration characteristic.The tests in this paper include no-load test and load test of transformer, short-circuit impact test of test transformer and on-line long-term monitoring test of 220kV power transformer, and a lot of research data have been obtained.The data analysis of off-line test shows that the short circuit impact has little effect on the vibration of iron core. After three rounds of short circuit impact, the frequency domain distribution of vibration signal changes little, and the standard deviation of frequency domain amplitude is 0.0711 / 0.077 / 8 / 0.0775, respectively.In this paper, the "frequency-energy-Euclidean distance" obtained by wavelet packet decomposition is used as the vibration eigenvalue to quantify the winding deformation.Taking phase A as an example, the data obtained from the load test are analyzed. After three rounds of impact tests, the characteristic values are 0.06480.1094 / 0.2062.In this paper, the on-line transformer vibration signal is separated by blind source separation algorithm. Through the analysis of simulation data and grid data, it is verified that the algorithm is suitable for transformer vibration signal separation.The neural network is used to train the neural network to recognize the vibration signal of the separated winding intelligently.The operation history of three 220kV transformers is statistically analyzed, and the vibration signals under the same load are compared and analyzed. It is verified that "frequency-energy-Euclidean distance" is also suitable for 220kV transformers as a vibration characteristic.But the diagnostic threshold is different from the test transformer.There are many shortcomings in the single data diagnosis, and the reliability and accuracy are not high.On the basis of detecting transformer winding deformation and vibration method, this paper uses a multi-information diagnosis model based on short-circuit reactance and vibration characteristic analysis to diagnose the state of transformer winding deformation. The result of data analysis proves that it has higher accuracy.It also has the ability to distinguish small deformation.
【学位授予单位】：沈阳工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM411

【相似文献】