基于GA-CNN-BiLSTM算法的变压器故障预测研究

Research on Transformer Fault Prediction Based on GA-CNN-BiLSTM Algorithm

摘要: 变压器是电力系统重要的设备，一旦发生故障可能会引发重大事故，须及时准确地发现其隐患。文章利用卷积神经网络融合双向长短期记忆网络建立了变压器故障预测模型。针对CNNBiLSTM算法丢失部分细节导致准确率不高的问题，采用遗传算法对CNNBiLSTM算法进行优化。实验结果表明，GA-CNN-BiLSTM在变压器故障预测中性能最佳，CNNBiLSTM次之，BiLSTM和支持向量机（support vector machine，SVM）的预测相对较低。因此，对于变压器故障预测，GACNNBiLSTM是一种更为有效可靠的方法。

Abstract: Transformer is an indispensable component of the power system. Once a fault occurs, it may lead to major accidents. It is necessary to find its fault timely and accurately. Therefore, a transformer fault prediction model is established by using convolutional neural network (CNN) and bidirectional long short term memory (BiLSTM) network. Aiming at the problem that the accuracy of CNN bilstm algorithm is not high due to the loss of some details, genetic algorithm (GA) is used to optimize CNN-BiLSTM algorithm. The experimental results show that GA-CNN-BiLSTM has the best performance in transformer fault prediction,CNN-BiLSTM takes the second place, the prediction of BiLSTM and support vector machine (SVM) is relatively low. Therefore, GA-CNN-BiLSTM is a more effective and reliable method for transformer fault prediction.