中文摘要：

针对变压器故障诊断中样本存在的随机性和模糊性,提出了一种基于Ada Boost优化云理论的变压器故障诊断方法。该方法通过使用云分解法对变压器油中溶解某气体体积分数占所有气体总体积分数的百分比及气体总体积分数（简称百分比及总量）为元素进行云分解,以不同元素的云概念相互组合构建云组合,并以云变换后的训练样本建立云组合到故障类型的隶属空间,得到云诊断模型。以该云诊断模型为基础分类器,通过Ada Boost算法对其迭代训练,获取一系列弱云诊断器,并利用Ada Boost算法的集成提升作用,由加权投票法产生强诊断器。结果表明：该方法所建立的云组合诊断正确率高于基于气体含量（体积分数）的云组合,通过Ada Boost算法能进一步提高诊断正确率,但该模型诊断正确率仍高于基于气体含量的云组合。研究证明基于百分比及总量的云分解法具有更优秀的分类能力,且通过Ada Boost修正集成能力,诊断能力会在保持快捷性的基础上进一步加强。

英文摘要：

In order to resolve the problem that the samples for power transformer fault diagnosis with certain randomness and fuzziness, we proposed a new fault diagnosis method based on the cloud model of AdaBoost algorithm. Using the cloud decomposition method, we obtained elements through analyzing the total dissolved gas contem and the percentage of dissolved gas in transformer oil. Then we built cloud combination with cloud conceptions of different elements, and established a new cloud fault diagnosis model based on the relation space between fault type and the cloud combination consisted of transformed training sample. Using this model as the basic classifier, a series of weak cloud models can be obtained through iterative training with AdaBoost algorithm. According to the integration effect of AdaBoost algorithm, we can establish a strong cloud model by a weighted voting system. Further case studies show that the diagnostic accura- cy of the proposed model is higher than the cloud model based on gas contents, while AdaBoost algorithm can further increase the recognition ratio and diagnostic accuracy of the proposed model. Moreover, the proposed cloud model is also better in classification capacity, which can be further improved by the amendment and integration effect of AdaBoost al- gorithm without affecting the model＇s efficiency.