基于改进量子粒子群优化模糊聚类的变压器故障诊断方法

Research on transformer fault diagnosis method based on improved QPSO-FCM algorithm

对电力变压器进行高效准确的故障诊断可有效保障电力系统安全、稳定运行。为提高变压器故障诊断正确率,提出了一种基于改进量子粒子群优化模糊聚类的变压器故障诊断方法。采用遗传算法杂交概率的思想改进量子粒子群算法提高算法收敛速度、防止陷入局部极值,克服模糊聚类算法易受初始值影响的不足,进而实现对变压器高效、准确的故障诊断。以变压器油中典型气体作为故障特征量,选取68组数据建立故障集,采用改进量子粒子群算法寻找最佳初始聚类中心,并将其应用于3种不同数据组进行验证,实验结果表明文中所提方法的有效性。

Having an efficient and accurate diagnosis for power transformers fault is a key aspect for the safety and sta- bility of the power system. To improve the accuracy rate of transformer fault diagnosis, a transformer fault diagnosis method based on improved quantum-behaved particle swarm optimization fuzzy clustering （IQPSO-FCM） is proposed. To overcome the shortcomings for sensitivity to initial value in fuzzy clustering method, the QPSO is improved by ge- netic hybrid algorithm to increase the convergence speed and prevent falling into local extremes. Thus; an efficient and rapid fault diagnosis result for transformers was obtained. In this article, the dissolved gas in oil was taken as the characteristic quantity of fault, the fault set was composed of 68 groups of data, and IQPSO was adopted to gain the optimal initial cluster centers for verifying the 3 different data sets. Experimental results show that the effectiveness of the proposed method.