基于改进果蝇算法优化PNN的变压器故障诊断研究

为了提高电力变压器故障诊断准确率,通过分析变压器油中溶解气体数据,提出了一种定向变步长的果蝇算法(DVSFOA)与概率神经网络(PNN)相结合的变压器故障诊断模型。由于PNN的参数平滑因子对输出结果影响较大,对果蝇算法位置公式进行更新调整,对平滑因子进行参数寻优,将优化结果赋值给PNN模型进行网络训练,得到了用于变压器故障诊断的最佳网络模型。实验结果表明,该组合算法具有较高的诊断精度,收敛速度快,整体性能高。

Transformer Fault Diagnosis Based on Improved FOA Optimized PNN

In order to improve the efficiency of power transformer fault diagnosis,and analyze the dissolved gas data in transformer oil,directional variable step fruit fly optimization algorithm (DVSFOA) combined with PNN for transformer fault diagnosis is proposed.Smoothing factor which is the parameter of PNN has great influence on the correctness of network output.In this paper,the position formula of FOA is updated and adjusted to find the optimal smoothing factor.The optimal network model for transformer fault diagnosis is obtained by assigning the optimal smoothing factor to PNN model for network training.The experimental results show that the combined algorithm has higher diagnostic accuracy and faster convergence speed,and the overall performance is high.