基于化学反应优化神经网络与融合DGA算法的油浸式变压器模型研究

Diagnosis Model for Transformer Fault Based on CRO-BP Neural Network and Fusion DGA Method

为了及时准确掌握变压器的健康状况,对潜伏性故障进行预测分析,将人工智能算法与DGA算法相结合,提出了一种基于化学反应优化神经网络的变压器故障诊断模型。考虑到BP神经网络和传统DGA算法在变压器故障诊断应用过程中存在的缺陷,在模型中引入化学反应优化算法和融合DGA算法对其进行改进。通过实例分析表明,提出的故障诊断模型的诊断准确率达到87.88%,迭代次数和训练时间分别为1991次和1927 ms;与其他诊断模型相比,模型在诊断效率和训练时间上具有明显的优势,对于变压器的故障预测和实时诊断具有一定的参考意义。

In order to comprehend the status of transformer and predict the development of the incipient fault, we put forward a transformer fault diagnosis model based on CRO-BP neural network, which combines both advantages of artificial intelligence and DGA method. In the proposed model, chemical reaction optimization and fusion DGA method are employed to overcome the defects of BP neural network and traditional DGA method. The fault diagnosis process of the model is presented in experiments and simulations.The results reveal that the accuracy, iterations, and training time of the model are 87.88%, 1 991, and 1 927 ms, respectively. Compared with those of other models, the results of the model demonstrate that the model has distinct advantages, and has reference significance in the prediction and real-time diagnosis of transformer faults.