中文摘要：

针对变压器故障综合诊断的问题，提出一种基于证据理论的故障综合诊断方法。为实现合成规则中冲突信息在合成后基本概率分配函数中的合理赋值，基于不同证据体对同一假设的紊乱度分配冲突信息，改进了传统的合成规则。将这一合成规则应用到基于粗糙集、模糊聚类、神经网络、贝叶斯理论等技术所形成证据体的合成之中。试验表明所提判断方法的适应性，在多种判断结果出现明显分歧的情况下，合理地处理了证据体之间的冲突信息，从而有效综合了各种不同方法的判断结果。

英文摘要：

For the problem of synthesized diagnosis of transformer faults, a method of synthesized diagnosis based on evidence theory is proposed. Firstly, for the rational assignment of conflict information to basic possibility assignment, the traditional combination rule is modified based on the degrees of disorder of each presumption. Secondly, the modified rule is used to combine the evidence bodies formed from data analysis technologies by rough set, fuzzy clustering, artificial neuron network and Bayes theory. Finally, the proposed method is used in a real example. The results of test show the feasibility of this proposed diagnosis method which can synthesize the diagnosis results determined by different diagnosis methods by using new combination rule to assign conflict information even if under the circumstance of notable differences in results.