基于PAS与DFNN的变压器故障预测研究

发布时间：2018-01-18 08:39

本文关键词：基于PAS与DFNN的变压器故障预测研究　出处：《河北联合大学》2014年硕士论文　论文类型：学位论文

【摘要】：变压器是电力系统和智能变电站中重要的电力设备，智能变压器智能化水平关系着智能变电站运行的可靠性和投资的经济性。而变压器故障预测能够发现潜伏的故障以及预告故障的发展趋势，研究故障预测对系统安全运行和变压器的状态检修有重要意义。动态模糊神经网络具有强大的多元非线性数据处理和函数逼近功能，能够利用原始样本数据通过模型内部自我学习训练获得准确度较高的预测诊断模型。将动态神经网络强大的预测诊断功能引入到变压器故障处理中，建立起能够真实反映变压器故障特性的智能预测诊断模型，能够实现变压器故障在线检测的要求，提高变电站的综合自动化水平。本课题结合变压器故障预测诊断在线监测的特点，选用了光声光谱技术对变压器油中的故障气体的含量进行实时在线的监测，，选取了动态模糊神经网络为实验的主要模型结构，利用MATLAB中的神经网络工具箱，建立起基于动态模糊神经网络的电力变压器故障预测模型。实验选取了150组变压器故障原始样本数据对D-FNN模型中进行学习训练，得到了具有预测诊断功能的网络模型；再挑选100组变压器的在线监测数据进行仿真试验，并查查看了模型预算误差收敛曲线，证明了采用基于PAS与DFNN变压器故障诊断预测模型预测变压器故障相对于传统的方法具有更高的故障诊断率,验证了基于PAS与DFNN在变压器故障预测诊断处理中的合理有效性。
[Abstract]:Transformer is an important power equipment in power system and intelligent substation. Intelligent transformer intelligence level relates to the reliability of intelligent substation operation and the economy of investment, and transformer fault prediction can find latent faults and forecast the development trend of fault. It is important to study the fault prediction for the safe operation of the system and the condition maintenance of the transformer. Dynamic fuzzy neural network has powerful functions of multivariate nonlinear data processing and function approximation. The predictive diagnosis model with high accuracy can be obtained by using the original sample data through self-learning training within the model. The powerful predictive diagnosis function of dynamic neural network is introduced into transformer fault processing. An intelligent predictive diagnosis model which can truly reflect the fault characteristics of transformers is established and the requirement of on-line detection of transformer faults can be realized. Based on the characteristics of on-line monitoring of transformer fault prediction and diagnosis, the photoacoustic spectrum technology is used to monitor the content of fault gas in transformer oil in real time and online. The dynamic fuzzy neural network is selected as the main model structure of the experiment, and the power transformer fault prediction model based on the dynamic fuzzy neural network is established by using the neural network toolbox in MATLAB. 150 sets of original transformer fault samples are selected for learning and training in D-FNN model, and a network model with predictive diagnosis function is obtained. Then the on-line monitoring data of 100 groups of transformers are selected for simulation test, and the convergence curve of model budget error is checked. It is proved that the prediction model of transformer fault based on PAS and DFNN has higher fault diagnosis rate than the traditional method. The validity of PAS and DFNN in transformer fault diagnosis and treatment is verified.
【学位授予单位】：河北联合大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM407

【参考文献】