基于马尔可夫转换场与多头注意力机制的电能质量扰动分类方法

Power Quality Disturbances Classification Based on Markov Transition Field and Multi-head Attention

新型电力系统中的电能质量扰动愈加复杂,为提升电能质量复杂扰动分类准确率并增强算法的噪声鲁棒性,提出了一种基于马尔可夫转换场与多头注意力机制的电能质量扰动分类方法。首先,利用马尔可夫转换场对电能质量扰动时序数据进行模态变换,得到图像模态数据;然后,将图像模态数据输入卷积神经网络进行特征提取;最后,利用多头注意力机制着重关注卷积神经网络提取特征的重要部分并进行扰动分类。与常规的图像模态转换方法相比,该方法具有更好的扰动分类效果与抗噪声能力。

Power quality disturbances in the new power systems have become more complex.In order to improve the classification accuracy of the complex power quality disturbances and enhance the noise robustness of the algorithm,a power quality disturbances classification based on the Markov transition field and the multi-head attention is proposed.Firstly,the Markov transition field is used to transform the power quality disturbances time sequence data,and the image modal data is obtained.Then,the image modal data is put into the convolutional neural network for the feature extraction.Finally,the multi-head attention is used to focus on the important part of the feature extraction of the convolutional neural networks and to classify the disturbances.Compared with the conventional image modal conversion,this method has better disturbance classification effect and anti-noise ability.