基于海鸥算法优化概率神经网络的变压器故障诊断

Transformer Fault Diagnosis Based on Seagull Algorithm Optimized Probabilistic Neural Networ

传统概率神经网络PNN依靠经验取平滑因子进行变压器故障诊断,导致预测准确率不高,因此提出一种基于无编码比值法(九维度)的海鸥算法优化概率神经网络PNN,构建一种新型变压器故障诊断模型。该方法用海鸥算法优化平滑因子,再将结果赋值给PNN模型进行训练,减少人为因素对神经网络设计的影响。结果表明,相对于传统PNN模型、PSO-PNN模型及SSA-PNN模型,该方法准确率更高,为变压器故障诊断提供了新方法。

The traditional probabilistic neural network(PNN)for transformer fault diagnosis relies on experience to get smoothing factor,which leads to low prediction accuracy.Therefore,a seagull algorithm based on uncoded ratio method(nine dimensions)is proposed to optimize PNN and a new transformer fault diagnosis model is built.The method uses seagull algorithm to optimize the smoothing factor,and then assigns the result to the PNN model for training,so as to reduce the influence of human factors on the design of neural network.The results show that compared with the traditional PNN model,PSO-PNN model and SSA-PNN model,the proposed method has higher accuracy and provides a new method for transformer fault diagnosis.