【摘要】 针对传统电能质量扰动识别和分类方法存在分类准确率低、泛化能力差、鲁棒性弱的问题，提出了一种基于深度学习的一维多尺度深度残差网络(1D-MDR)对电能质量扰动进行识别分类的方法。该方法首先引入多尺度熵对扰动信号进行预处理，有效地表征扰动信号的多尺度特征；然后将多尺度特征信号输入到基于自适应软阈值的深度残差网络中，对多尺度电能质量扰动信号进行特征提取并融合；最后，利用全连接网络实现电能质量扰动信号的分类。仿真实验表明，提出的方法能够自动、准确地进行特征提取，并进行有效的识别分类。通过对比实验，证明该方法具有更高的准确率。更多还原

【Abstract】 Aiming at the problems of low classification accuracy, poor generalization ability, and weak robustness in the traditional power quality disturbance identification and classification methods, a one-dimensional multi-scale deep residual convolutional neural network structure(1 D-MDR)is constructed and the identification and classification method of power quality disturbance is proposed based on the network.For the method, the multi-scale coarse-grained operations is firstly used to perform multi-scale feature characterization of the disturbance signals.And then the multi-scale characteristic signals are inputted into a deep residual network based on an adaptive soft threshold so as to extract and fuse the features of the power quality disturbance signals.Finally, the fully connected layer and multiple classification functions are utilized to realize the classification of power quality disturbance signals.A great number of experimental results show that by the method, the characteristic information of the disturbance signals can be accurately extracted and the effective recognition and classification can be performed.The comparative experiments prove that the method has a higher accuracy rate.更多还原