基于神经网络的变压器故障诊断系统的设计与实现

发布时间：2018-05-31 22:45

本文选题：神经网络 + BP算法　；参考：《郑州大学》2014年硕士论文

【摘要】：变压器是电力系统核心的部件之一，，担负着电压变换的重要任务，它的正常工作是整个电力系统正常供电的必要条件。如果变压器发生了故障，那么将对整个电力系统正常供电造成极大的破坏。因此，能够快速、准确的诊断出变压器故障类型对变压器故障的及时、迅速维修复原具有重要的意义。而神经网络尤其是BP神经网络具有结构简单、并行性高、非线性关系建模能力强等特点，非常适合用来解决故障诊断这类多变量的、内部关系复杂的问题。所以，将神经网络应用于变压器的故障诊断是可行的。许继集团是我国电力装备研发和电力生产的大型骨干企业，一直都很重视变压器故障诊断技术的发展和研究，近几年与我校老师合作研发变压器故障诊断系统。基于该项目的需求，本文做了两方面工作：一是研究神经网络相关理论算法，并将神经网络理论应用于变压器故障诊断，这方面主要工作是掌握BP算法以及应用BP神经网络进行应用建模；二是设计并在VC++6.0的平台上实现了基于神经网络的变压器故障诊断系统，该系统包含三个模块：数据管理、网络训练和故障诊断。数据管理模块主要实现了故障样本的预处理，网络训练模块主要实现了三种BP算法来训练网络，故障诊断模块主要实现对故障样本的诊断。在本文中，我们以来源于实际生产环境的60组变压器故障数据为例展示了本系统的效果，使用L-M算法诊断故障准确率最高可以达到91%。同时，本系统在许继集团的相关产品中得到了实际应用，满足了用户的需求。
[Abstract]:The transformer is one of the core components of the power system, which takes on the important task of the voltage transformation. Its normal work is the necessary condition for the normal power supply of the whole power system. If the transformer fails, it will cause great damage to the normal power supply of the whole power system. Type is of great significance to the timely and rapid maintenance and restoration of transformer faults. The neural network, especially the BP neural network, has the characteristics of simple structure, high parallelism and strong nonlinear relation modeling ability. It is very suitable to solve the problem of the multivariable and complex internal relations of fault diagnosis. So, the neural network is applied to the neural network. The fault diagnosis of the transformer is feasible.
Xu Ji group is a large backbone enterprise of power equipment R & D and power production in China. It has always attached great importance to the development and research of transformer fault diagnosis technology. In recent years, the transformer fault diagnosis system has been developed in cooperation with our teachers. Based on the requirements of this project, this paper has done two aspects of work: one is to study the theory of neural network related theory. The neural network theory is applied to the transformer fault diagnosis. The main work is to master the BP algorithm and apply the BP neural network to model the application. Two is to design and implement the transformer fault diagnosis system based on the neural network on the VC++6.0 platform. The system contains three modules: data management, network training and the system. The data management module mainly realizes the preprocessing of the fault samples. The network training module mainly implements three kinds of BP algorithms to train the network, and the fault diagnosis module mainly realizes the diagnosis of the fault samples.
In this paper, 60 sets of transformer fault data from the actual production environment have been presented as an example to show the effect of this system. The L-M algorithm is used to diagnose the highest fault accuracy and can reach the highest level of 91%.. The system has been applied to the related products of Xu Ji group and meets the needs of the users.
【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM407

【参考文献】