针对变压器故障的特征,结合变压器油中气体分析法以及三比值法.提出了基于遗传算法改进极限学习机的故障诊断方法。由于输入层与隐含层的权值和阈值是随机产生.传统的极限学习机可能会使隐含层节点过多,训练过程中容易产生过拟合现象。该方法运用遗传算法对极限学习机的输入层与隐含层的权值与阈值进行优化,从而提高模型的稳定性和预测精度。将诊断结果与传统的基于极限学习机故障诊断进行对比,结果表明,基于遗传算法改进极限学习机变压器故障诊断的精度更高。
Aiming at the characteristics of transformer fault types,a fault diagnosis method based on the improved extreme learning machine with genetic algorithm is proposed by combining with the transformer oil-dissolved gas analysis and the 3-ratio method. Because the weights and thresholds between input layer and hidden layer are generated randomly,the traditional extreme learning machine may generate excessive hidden layer nodes,resulting in over-fitting in the training process. Therefore, in the proposed method genetic algorithm is adopted to optimize weights and thresholds between input layer and hidden layer of extreme learning machine for improving the stability and prediction accuracy of extreme learning machine. Compared with the conventional fault diagnostic method based on extreme learning machine,the proposed fault diagnosis method of transformer achieves higher diagnosis precision.