基于极限学习机的变压器故障诊断

发布时间：2018-11-28 16:13

【摘要】：使用对有种溶解气体分析的方法进行变压器故障诊断，可在变压器运行期进行故障分析的特点，对于变压器维修模式的转变有很大的推动作用，具有重要的研究意义。本文在分析现有变压器故障诊断方法的特点及其存在问题的基础上，将极限学习机算法应用于变压器故障诊断。提出了基于极限学习机的油浸式电力变压器故障诊断方法。分析了不同隐藏层激活函数对极限学习机的诊断性能的影响，给出了诊断的具体实现方法。这种方法有不容易出现局部值的特点，且训练速度快，参数设定简单，易于应用，，适合于在线诊断。并通过实例验证了该方法的性能。给了基于WELM的变压器故障诊断方法。这种方法主要针对DGA数据中存在的数据不均现象，使用加权方案使数据恢复平衡性。研究了不同加权方案对诊断性能的影响。通过实验证明了WELM有更好的诊断效果。在研究KELM参数优化的基础上提出了基于KELM的变压器故障诊断方法。提出了使用粒子群优化算法结合K折交叉验证的方法对KELM参数进行优化的方法，给出了具体参数优化和诊断实现过程。实验证明，相比SVM算法，基于KELM的变压器故障诊断方法诊断准确率更高，训练时间更短。
[Abstract]:Using the method of dissolved gas analysis for transformer fault diagnosis, it can be used to analyze the characteristics of transformer fault during the operation period, which has a great role in promoting the transformation of transformer maintenance mode, and has an important significance in research. On the basis of analyzing the characteristics of existing transformer fault diagnosis methods and their existing problems, this paper applies the extreme learning machine algorithm to transformer fault diagnosis. An oil-immersed power transformer fault diagnosis method based on extreme learning machine is proposed. The influence of different hidden layer activation functions on the diagnostic performance of LLM is analyzed, and the realization method of diagnosis is given. This method is not easy to appear the local value, and the training speed is fast, the parameter setting is simple, the method is easy to be applied, and it is suitable for on-line diagnosis. The performance of the method is verified by an example. The method of transformer fault diagnosis based on WELM is given. This method mainly aims at the uneven data in DGA data, and uses the weighted scheme to restore the balance of the data. The influence of different weighting schemes on diagnostic performance was studied. Experimental results show that WELM has better diagnostic effect. Based on the study of KELM parameter optimization, a transformer fault diagnosis method based on KELM is proposed. The particle swarm optimization (PSO) algorithm combined with K-fold cross-validation is proposed to optimize KELM parameters. The process of parameter optimization and diagnosis is given. Experimental results show that compared with SVM algorithm, transformer fault diagnosis method based on KELM has higher accuracy and shorter training time.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM407

【相似文献】