基于概率神经网络的小电流接地系统模式识别故障选线方法及应用

发布时间：2018-12-08 18:20

【摘要】：小电流接地系统是指中低压配电网络中，，采用中性点经消弧线圈接地、经大电阻接地或者中性点不接地等运行方式的系统。通常情况下，小电流系统发生故障后，故障电流十分微弱，不容易被检测到，这对故障选线问题提出了巨大挑战。目前小电流接地故障选线方法主要分为三类：一是利用故障稳态特征分量进行选线，二是利用故障暂态特征分量进行选线，三是注入特殊信号进行选线。这些方法在实际应用中均取得了一些成效，但从选线的准确率和稳定性上来说还远远不够理想。概率神经网络是一种可用于模式分类的神经网络，在机械、材料、环境工程乃至经济领域有着较好应用，而在配电网故障选线上很少有人尝试。经过反复研究和探索，寻找到配电线路零序电流小波能量、有功分量和五次谐波分量三种故障特征量作为选线的依据。对故障模式进行了合理定义，重点突破了概率神经网络中多种故障特征量有效融合的问题，提出一种基于概率神经网络的模式识别选线方法。通过对小电流接地系统模型进行大量Matlab仿真试验，研究了故障位置、接地电阻、故障合闸角、中性点接地方式、配电线路结构和噪声干扰对故障特征量的影响，收集了大量故障数据样本。同时对电弧高阻接地、混合线缆配电网、噪声干扰下的接地故障以及中性点不同接地方式系统的概率神经网络模式识别选线进行了广泛试验，验证了该方法具有较好的通用性和抗干扰能力。将该方法与概率神经网络单一特征量选线和BP神经网络故障选线方法进行了比较，证明了该方法具有准确率高，操作简单快速，故障知识丰富，易于拓展知识库等特点。最后提出了一套本方法的故障选线装置设计方案。
[Abstract]:Low current grounding system is a system in which neutral point is grounded by arc-suppression coil grounding through large resistance or neutral point is not grounded in medium and low voltage distribution network. Usually, the fault current is very weak after the failure of the small current system, and it is not easy to detect, which poses a great challenge to the fault line selection problem. At present, there are three kinds of fault line selection methods for low current grounding fault: one is to select the line by using the steady-state characteristic component of the fault, the other is to select the line by using the transient characteristic component of the fault, and the third is to inject special signals to select the line. These methods have achieved some results in practical application, but they are far from ideal in terms of accuracy and stability of line selection. Probabilistic neural network (PNN) is a kind of neural network which can be used for pattern classification. It has good applications in mechanical, material, environmental engineering and even economic fields, but it is seldom tried in fault line selection of distribution network. Through repeated research and exploration, three fault characteristic quantities of zero-sequence current wavelet energy, active power component and fifth harmonic component of distribution line are found as the basis of line selection. In this paper, the fault mode is defined reasonably, and the problem of effective fusion of multiple fault features in probabilistic neural network is broken through, and a method of pattern recognition and line selection based on probabilistic neural network is proposed. Through a large number of Matlab simulation tests on the small current grounding system model, the effects of fault location, grounding resistance, fault closing angle, neutral grounding mode, distribution line structure and noise interference on the fault characteristic quantity are studied. A large number of fault data samples were collected. At the same time, a wide range of experiments have been carried out on arc high resistance grounding, hybrid cable distribution network, grounding fault under noise interference and probabilistic neural network pattern recognition of different neutral grounding mode systems. It is proved that this method has good generality and anti-interference ability. The method is compared with the probabilistic neural network single feature selection method and the BP neural network fault line selection method. It is proved that this method has the advantages of high accuracy, simple and fast operation, rich fault knowledge and easy to expand the knowledge base. Finally, a design scheme of fault line selection device based on this method is put forward.
【学位授予单位】：南昌大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM862

【参考文献】