基于粒子群优化支持向量机的变压器故障诊断

为了克服了人工神经网络(ANN)中存在的过拟合、收敛速度慢、容易陷入局部极值等缺点,提出了基于粒子群优化支持向量机(PSO-SVM)的变压器故障诊断方法,即将粒子群优化算法(PSO)用于SVM参数优化。PSO是一种智能群体搜索方法,它源于对鸟类捕食行为的研究。这种方法不仅具有很强的全局搜索能力,而且容易实现,适合于SVM参数优化。变压器故障诊断实例分析结果证明,PSO-SVM的诊断精度高于IEC三比值法、BP神经网络、普通的SVM,PSO-SVM适用于电力变压器故障诊断。

Fault Diagnosis of Transformer Based on Particle Swarm Optimization-based Support Vector Machine

Accurate assessment of system reliability with limited or insufficient statistical data is difficult. At present, to assess the reliability of the transformers primarily depend on the statistics data, and the result of the assessment does not accurately reflect the reliability of a transformer. Fault tree is a powerful tool for depicting the logical relationships between the faults system. However, for a sophisticated reliability system, like power transformer,fault analysis is hard or infeasible because of the complexities of the equipment configuration and its faults. This paper presents a method which overcomes the drawbacks of traditional fault-tree analysis (FTA) by using FTA based on fault probability and equivalent calculation of each failure mode. Based on the further analysis of typical failure modes and influences of power transformer, the fault-tree of transformer is set up by using FTA techniques and the assessment method for the reliability of transformer is put forward, combined with the index of the importance and the method of improved analytic hierarchy process(IAHP). The assessment examples show the method can deduct the failure modes of transformer clearly, meanwhile, find out the weaknesses in time during the operation and get better assessment results. The method is helpful for the reliability analysis of a sophisticated system.