摘要
为了实现对变压器油电气性能快速准确地检测,提出了基于多频超声波(MFU)技术和灰狼优化BP神经网络(GWO⁃BP)的检测方法,利用不同频率超声波在变压器油中传播速度和衰减系数不同、对不同尺寸颗粒物检测效果不同的特点,综合了超声波在无损检测上的优势以及灰狼算法全局寻优能力强、预测精度高的优点。首先利用超声波在非均匀介质中的传播特性,准确地获取超声波在变压器油中的传播速度、衰减系数等特性参数;再利用灰狼算法对传统BP神经网络进行优化。然后利用灰狼优化的神经网络对样本进行训练,建立超声波传播特性参数和变压器油电气性能参数之间的映射关系,进而建立变压器油电气性能参数的检测模型,实现对多个电气性能参数快速准确的无损检测。通过实验,验证了该方法在检测变压器油电气性能参数上的可行性和有效性。
To realize rapid and accurate detection of transformer oil’s electrical performance,a detection method based on the multi⁃frequency ultrasonic(MFU)technology and the grey wolf⁃optimized BP neural network(GWO⁃BP)is proposed.This method makes use of the characteristics that the ultrasonic wave with different frequency has different propagation speed and attenuation coefficient in transformer oil,hence has different particle size detection effect.An this method combines the advantages of ultrasonic wave in nondestructive testing,the advantages of multi⁃frequency ultrasonic technology in fast detection speed and comprehensive information acquisition,and the advantag⁃es of grey wolf⁃optimizer in global optimization and high prediction accuracy.Firstly,the propagation velocity and at⁃tenuation coefficient of ultrasonic wave in transformer oil are accurately obtained by using the propagation character⁃istics of ultrasonic wave in inhomogeneous medium.The grey wolf⁃optimizer is used to optimize the traditional BP neural network.Then,the grey wolf⁃optimized neural network is used to train the samples to establish the mapping re⁃lationship between the ultrasonic propagation characteristic parameters and the transformer oil’s electrical perfor⁃mance parameters.Furthermore,the detection model of the electrical performance parameters is established to real⁃ize fast and accurate nondestructive testing of multiple electrical performance parameters.The feasibility and effec⁃tiveness of the method in detecting the electrical performance parameters of transformer oil are verified by experi⁃ments.
作者
王旭
生西奎
慕锟
王昕
WANG Xu;SHENG Xikui;MU Kun;WANG Xin(Center of Electrical&Electronic Technology,Shanghai Jiao Tong University,Shanghai 200240,China;Yanbian Power Supply Company,Jilin Electric Power Company,State Grid Corporation of China,Jilin Yanbian 133000,China)
出处
《高压电器》
CAS
CSCD
北大核心
2020年第8期114-120,共7页
High Voltage Apparatus
基金
国家自然科学基金(61673268,61533012)
国网吉林省电力有限公司2019年科技立项(2019⁃13)。
关键词
变压器油
电气性能参数
多频超声波
灰狼算法
BP神经网络
transformer oil
electrical performance parameters
multi⁃frequency ultrasonic
grey wolf⁃optimizer
BP neural network
