变电站过电压仿真分析及分层识别研究

发布时间：2018-03-06 15:04

本文选题：过电压　切入点：ATP-EMTP　出处：《西安理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：电力系统出现扰动或故障时产生的过电压是危害电网安全运行的重要因素之一,过电压可毁坏电气设备,导致供电中断甚至大范围停电等严重事故,不同类型的过电压有着不同的产生原理,其防范措施也各不相同。因此,如何准确、迅速的识别出不同类型的过电压,对及时排除故障、保障电力系统稳定安全运行有着重大意义。本文根据某变电站现场实际接线及相关电气设备具体参数,对线路、变压器、电源、负荷、避雷器等主要设备在ATP-EMTP仿真软件中进行模型选择及参数计算,完成设备参数设置并搭建该变电站过电压仿真模型,对电容器组合闸、空载线路合闸、空载变压器合闸、单相接地、直击雷以及感应雷过电压等6种常见的典型过电压波形进行了仿真计算,结合各自过电压产生原理,分析并总结了每一过电压波形特点。根据仿真计算结果,进行了变电站过电压特征量提取的研究。为了对不同等级过电压波形进行比较,对其进行了归一化处理。依据不同过电压持续时间特点,对仿真得到的波形区间划分为3个时间段,在此基础上,根据不同波形间的差异选取不同区间,采用时域分析法提取了 6个相对明显的特征量,对直观上不易区分的波形,采取小波分析法提取了8个具有较好差异性的特征量,并利用仿真数据验证了该14个特征量差异的有效性。最后,综合仿真得出的6种变电站过电压波形特征,构建了基于支持向量机的过电压分层识别结构并设计识别流程,可由上到下对所提取的特征量在每层分类器中进行选择。为验证该识别系统的有效性,根据仿真得到480条过电压数据,240条用于训练,240条用于测试。结果表明,本文所采用的支持向量机的识别准确率接近88%,识别时间近乎2.5s,相比传统识别方法准确率提升约12%,时间缩短约3s,该方法准确性高,速度快,能够有效对不同类型过电压进行识别,为今后防止过电压的工程应用中提供技术支持。
[Abstract]:Power system disturbance or fault overvoltage produced is one of the important factors endangering the safety of power grid, overvoltage can destroy the electrical equipment, causing power outages or blackout accident, overvoltage of different type has different principle, the prevention measures are also different. Therefore, how to accurately quickly, identify different types of overvoltage, timely troubleshooting, is of great significance to ensure the safe and stable operation of power system. According to the actual substation wiring and electrical equipment related to specific parameters of the circuit, transformer, power supply, load, main equipment of lightning arrester for model selection and the parameter in the ATP-EMTP simulation software in the calculation, complete set of device parameters and build the substation overvoltage simulation model of capacitor combination switch, no-load circuit switch, transformer no-load switch, single-phase grounding, Direct lightning and lightning over-voltage to 6 kinds of typical voltage waveform is simulated, with their respective overvoltage principle, analyzes and summarizes each waveform characteristics. Based on the simulation results, the research on substation overvoltage feature extraction. In order to compare the different grade voltage waveform and the normalized. According to different voltage duration characteristics of waveform interval division by simulation for the 3 time, on this basis, according to the difference between the different waveform interval, time-domain extraction of 6 relatively obvious characteristic analysis method is used to visually indistinguishable take the 8 waveform extraction has better difference characteristics of wavelet analysis, and using the simulation data to verify the validity of the 14 characteristics of volume differences. Finally, 6 comprehensive results Substation overvoltage waveform characteristics, construct the over-voltage layeredidentification structure design and process identification based on support vector machine, from top to bottom on the features extracted in each layer of the classifier selection. To verify the effectiveness of the recognition system, according to the simulated 480 overvoltage data, 240 for training 240, for testing. The results show that the support vector machine recognition the recognition accuracy rate close to 88%, almost 2.5s time, compared with the traditional identification methods to enhance the accuracy of about 12%, shorten the time of about 3S, the method is of high accuracy, fast speed, can effectively identify the different types of overvoltage, to prevent future provide technical support for engineering application of voltage.

【学位授予单位】：西安理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM63

【参考文献】