盲源分离在电气设备温度在线监测中的应用

为了能分离出电气设备在线监测中的温度场图像中的温度信号和噪声信号,给出了一种基于自然梯度算法的盲源分离方法。在简单阐述了盲源分离的基本原理,介绍了盲源分离的基本模型、互信息最小化判据和自然梯度算法后,给出了所提方法的具体实现步骤、仿真实验结果及实际温度场图像的处理结果。结果表明利用该方法可以有效去除了温度场图像中的噪声,提高了在线监测数据的稳定性和可靠性,具有工程应用价值。

Blind Source Separation and Its Application in Temperature On-line Monitoring for Electrical Equipment

In order to separate temperature signal and noise signal from the temperature field image of on-line monitoring electric equipment,this paper provides a method to separate the blind source,which is based on the natural gradient algorithm.First,the paper describes the problem of blind signal separation.And,the principle and the linear model of BSS are introduced briefly.Then,the criterion of mutual information minimization is discussed.Its basic task is to optimize the separation matrix.The natural gradient algorithm is presented based on this criterion.It is an on-line adaptive algorithm and can avoid the matrix inversion so that its computational complexity is low,speed of convergence is fast,and stability is good.Finally,the detailed processes of the method and the results are presented.For the sake of proving the validity of natural gradient algorithm,the simulation experiment is given.Also,the detailed steps of temperature field image de-noising processing are proposed.The results show that it can remove the noise from the image effectively,which improves the stability and reliability of monitoring data.Moreover,it is proved that the method has the engineering practice value.