基于广义回归神经网络的变压器振动信号预测  

Application of Generalized Regression Neural Network in Predicting the Vibration Signal of Power Transformer

作　　者：林嘉扬[1] 

机构地区：[1]上海市电力公司超高压输变电公司,上海200057

出　　处：《华中电力》2012年第3期95-98,共3页Central China Electric Power

摘　　要：变压器箱壁的振动信号与绕组和铁芯的运行状态密切相关。在对变压器振动机理进行分析的基础上,提出使用广义回归神经网络获取变压器电压电流信号与振动信号100Hz分量的对应关系,据此对变压器振动信号进行预测。使用模型变压器试验平台获取的电压电流和振动信号对广义回归神经网络进行训练,并把训练好的神经网络用于振动信号的预测。预测信号与实测信号的良好吻合说明了文中所提出的方法的有效性,据此可对变压器运行状态进行在线监测。The vibration signals of power transformer are closely related to the operation conditions of winding and core.Based on the analysis of vibration mechanism of power transformer,the generalized regression neural network is applied to obtain the relations between the 100Hz component and voltage signal and current signal of power transformer,then the vibration signal can be predicted accordingly.The experiment platform is built to obtain the voltage,current and vibration signal to test the generalized regression neural network and to predict the vibration signal with the obtained neural network.The predicted vibration signals are fit well with the practical vibration signal,which verifies the effectiveness of the proposed method.Then the online monitoring of operation condition of power transformer could be easily realized.

关 键 词：变压器 振动 广义回归神经网络 在线监测 

分 类 号：TM401+.1[电气工程—电器]