ABSTRACT

Classification of faults in transformers with high accuracy is fundamental to ensuring good power quality with least interruptions. Our current work develops an intelligent genetic algorithm (GA)-tuned fuzzy classifier for transformer fault identification. The proposed classifier is able to segregate all fault types using dissolved gas analysis (DGA) samples from real power transformers of HPSEB (India) and other sources. DGA samples have been pre-processed using the J48 algorithm. We propose to replace the conventional action selection procedure of reinforcement learning by a GA-based optimizer. The classifier is able to garner very high classification accuracy which is higher than the one obtained with benchmark fuzzy Q learning (FQL) and other conventional classifiers. With our approach, the average fault classification rate achieved is 91.85% (FQL) and 97.51% genetic algorithm fuzzy Q-learning (GAFQL) though with a slightly higher computational complexity over the FQL. Our proposed classifier could serve as an important tool in ensuring the healthy operation of power transformers.

摘要

高精度的变压器故障分类是确保良好电能质量和最少中断的基础。我们目前的工作开发了一种智能遗传算法 (GA) 调整的模糊分类器，用于变压器故障识别。所提出的分类器能够使用来自 HPSEB（印度）和其他来源的真实电力变压器的溶解气体分析 (DGA) 样本来分离所有故障类型。DGA 样本已使用 J48 算法进行了预处理。我们建议用基于 GA 的优化器代替传统的强化学习动作选择过程。该分类器能够获得非常高的分类精度，高于使用基准模糊 Q 学习 (FQL) 和其他常规分类器获得的分类精度。使用我们的方法，实现的平均故障分类率为 91。85% (FQL) 和 97.51% 遗传算法模糊 Q 学习 (GAFQL) 虽然计算复杂度比 FQL 略高。我们提出的分类器可以作为确保电力变压器健康运行的重要工具。