摘要
由于电力变压器的故障原因、故障现象和故障机理间存在着随机性和模糊性等不确定因素,故障特征量和故障征兆间没有明确的一一对应关系,为了用特征量的数值大小来衡量和确定故障发生的原因和严重程度,在深入对变压器油中溶解气体分析的基础上,按照聚类分析的思想,运用模糊C-均值聚类分析原理(Fuzzy C-Means,FCM)和IEC三比值法,将这两种故障类型识别的方法在特征气体诊断分析中进行有效结合,建立了多诊断方法的系统框架。该法可以定量地确定故障在特征气体空间上的聚集效应,从而达到对故障进行自动合理分类的目的。仿真得出该方法正判率为98.1%,较IEC比值法、FCM算法的正判率分别提高了3.6%、4.5%。结果表明该方法具有很高的可行性和有效性。
Power transformer is one of the most important devices in electric power system. And monitoring the components of dissolved gas-in-oil can discover its operation problem, thereby, avoid availably electric power grid running accidents caused by power transformer fault. To use dissolved gas analysis （DGA） in oil can detect the po- tential fault of power transformer. As uncertain factors such as fuzziness and random exist among fault causes, fault phenomenon and fault type of power transformer, no correct one to one corresponding relations indwell in fault char- acteristic quantity and symptom of fault. Thus characteristic quantity cannot measure and identify fault causes and severity. So far one of main methods for transformer fault diagnosis is IEC three ratio codes. Though it has the defect of criterion lacking for transformer failure codes on critical value and other problems, it becomes the most common used method of current preventive test for power transformer. However, due to the problem of seldom interaction between IEC three ratio codes and other methods to diagnose transformer such as neural network, fuzzy expert system, genetic algorithm and artificial intelligence technique, most advantage and characteristics of IEC three ratio codes can not be revealed from these ways. According to the ideal of cluster analysis, Fuzzy C-Means （FCM） combined with IEC three-ratio method is adopted based on thorough analysis of dissolved gas in transformer to quantificational depict aggregation effect of characteristic gas space, which can categorize faults automatically and reasonably and also enhance the veracity and reliability of diagnosis with conventional IEC three ratio method in power transformer. Simulation results show the accurate judgement rate obtained using this new method is 98.1% which is higher than IEC three ratio codes and FCM methods by 6. 3% and 4. 5% respectively. Results show the high feasibility and effectiveness of this way.
出处
《高电压技术》
EI
CAS
CSCD
北大核心
2007年第8期66-70,共5页
High Voltage Engineering
基金
香港创新科技基金(ITS/054/06)。~~
