【作者】 赵磊；

【导师】 宋吉江；

【作者基本信息】 山东理工大学 ， 电气工程， 2016， 硕士

【摘要】 配电网系统在发生单相接地故障时会在全网络呈现零序电压,同时正常相的电压升高为线电压。规程规定该故障允许短时运行1~2小时,但随着配电网自动化建设的不断深入,配电网结构也日趋复杂,该故障对配电网系统的安全稳定运行所带来的隐患亟待解决。本文在总结相关专家学者的研究成果基础上,提出通过模糊神经网络理论融合稳态和暂态选线判据下的多种方案来解决小电流接地系统单相接地故障选线的难题。本文所做的主要工作如下:(1)总结了国内外相关专家学者针对小电流接地系统单相接地故障选线问题而提出的各种选线方法,并对已有的选线方法进行了评价,进而指出故障选线的一种发展趋势是基于人工智能理论下的稳态与暂态多判据融合方法。(2)对中性点不接地系统及谐振接地系统发生单相接地故障后零序电流的稳态和暂态特征进行了理论分析。(3)研究了基于模糊理论的多判据融合故障选线方案。对稳态选线判据中的零序电流五次谐波比相法、零序电流有功分量比幅法和暂态选线判据中的小波包系数极性法、暂态能量法的故障选线原理进行了详细说明,在此基础上建立了相应的故障测度隶属度函数。在不同的故障情况下对上述四种方法进行仿真分析,结果表明这四种方法的选线效果具有一定的优势互补特点,从而证明了对这四种选线方法进行融合的合理性。(4)对基于BP神经网络理论的故障选线方案进行了分析研究,指出了BP神经网络的不足,并采用遗传算法(Genetic Algorithm,GA)对BP神经网络的初始权值和阈值进行优化。分析了模糊理论和神经网络理论融合多判据进行故障选线的优缺点,并在此基础上提出了基于模糊理论和GA-BP神经网络理论相结合的多判据融合选线方法。(5)对本文提出的基于模糊GA-BP神经网络理论的多判据融合故障选线方法进行了仿真试验,仿真试验结果表明该方法具有较好的故障选线效果,从而为故障选线问题的解决提供了一定的思路和方法。对比了模糊GA-BP神经网络与模糊BP神经网络的选线效果,结果表明前者比后者具有更好的选线准确性,进一步证明了遗传算法对BP神经网络具有很好的改良效果。更多还原

【Abstract】 Distribution network system will present the zero sequence voltage in the single-phase grounding fault, while the normal phase voltage increases as the line voltage. Engineering allows the fault to operate in a short time 1~2 hours, but with the construction of power distribution automation, its structure is becoming more and more complex. The hidden dangers of the safe and stable operation need to be solved urgently. Based on the summary of relevant experts and scholars studies, the fuzzy neural network theory are proposed by compromising steady fault line selection methods and transient fault line selection methods to solve single-phase grounding fault line selection problem about non-effectively earthed system. The main works are as follows:(1)The fault line selection theories proposed by the relevant experts and scholars at home and abroad of single-phase grounding fault about the non-effectively earthed system are summarized. Existing fault line selection methods are evaluated. Then, it points out that the development trend of fault line selection methods is the combination of steady and transient fault line selection criteria by the theory of artificial intelligence.(2)The characteristics of steady and transient zero sequence current are analyzed theoretically in isolated neutral system and resonance grounding system about single-phase grounding fault.(3)A multi methods fusion fault line selection based on fuzzy theory is studied. Some fault line selection methods about zero sequence current five times harmonic specific phase method, active component of zero sequence current amplitude about the theory of steady-state line selection criterions and wavelet packet coefficient polarity method, transient energy method about the theory of transient line selection criterions are researched, then the corresponding fault measure membership functions is established. Finally, these four methods are proved to have complementary advantages and disadvantages by the simulation analysis under different fault conditions, which proves the reasonability of the integration for the above methods.(4)A fault line selection method based on BP neural network theory is analyzed and studied, the deficiency of neural network is pointed out. Then, the solution to the optimization of BP neural network by genetic algorithm is used. Analysis the advantages and disadvantages about the fuzzy theory and neural network theory. Then, a new fault selection method was proposed based on the fuzzy theory and GA-BP neural network to fuse multi methods.(5)The methods of fuzzy GA-BP neural network is simulated. The simulation results show that the method is effective. So as to provide an idea and method to solve the problem of fault line selection. Compared with the fuzzy GA-BP neural network and fuzzy BP neural network. The results show that the former has better accuracy than the latter and the genetic algorithm has a good improvement effect to BP neural network.更多还原