摘要
基于变压器油中溶解气体分析(DGA)法是使用神经网络和灰色预测对变压器的故障进行预测的。主要是采集变压器油在各种情况下的数据,并对应其故障进行编码,再用Matlab编写神经网络进行训练,输入各特征气体百分含量,输出对应的故障编码。通过对比,发现神经网络预测精度高达80%,使用灰色理论对各特征气体含量进行预测,与实际值对比,预测精度很高。最后将各个特征气体含量转化为百分数,输入已训练好的神经网络系统,预测出变压器的状态。最终所预测出的故障和实际故障一致。
Based on the dissolved gases atomizer （DGA） in transformer oil, combines the neural network and the gray theory to predict the transformer faults. Collects the data of transformer oil in different kinds of situation, and codes with the relevant fault. Then uses the soft of Matlab to program the neural network and then train it with the coded data. Inputs the percentage of the different dissolved gases, outputs the relevant fault code. The compare results find that the predict accuracy of neural network can be 80%. Then uses the gray theory to predict the amount of dissolved gases in transformer oil. Finds that the predictions are very close to the actual values. Follows, changes the amount of gas into percentage and then inputs the data into the well trained neural network to predict the fault. The prediction situation is same as the actual situation.
出处
《煤矿机电》
2016年第6期17-22,共6页
Colliery Mechanical & Electrical Technology
