一种新型用于电能质量扰动信号分类的混合深度学习方法

A Novel Hybrid Deep Learning Method for Power Quality Disturbance Classification

摘要: 由于电力电子设备的广泛使用和电力资源的过度消耗，节约能源迫在眉睫。电力系统中的非线性负载和其他负载被认为是电能质量扰动的主要原因，电能质量扰动的产生会导致的信号质量和形状的失真，从而导致总效率的降低。于是，提出一种新型的混合卷积神经网络方法由一维卷积神经网络结构和二维卷积神经网络组成，用来分类电能质量扰动信号。这两种卷积神经网络架构所获得的特征使用全连接层进行分类，功率信号使用原始形式的一维卷积神经网络进行处理。然后将这些信号转换为图像并使用二维卷积神经网络处理，结合一维和二维卷积神经网络生成的特征向量。最后，通过完全连接的层对该组合向量进行分类。所提出的方法非常适合信号处理的性质。这是一种新颖的方法。将所提出的框架与文献中其他最先进的电能质量扰动分类方法进行了比较。虽然与其他方法相比，所提出的方法的分类性能相对较高，但计算复杂度几乎相同。

Abstract: Due to the widespread use of power electronic devices and the excessive consumption of power resources, saving energy is imminent. Nonlinear loads and other loads in power systems are considered to be the main cause of power quality disturbances, which can lead to distortion of signal quality and shape, resulting in a reduction in overall efficiency. Therefore, a novel hybrid convolutional neural network method is proposed, which consists of a 1D convolutional neural network structure and a 2D convolutional neural network, and is used to classify power quality disturbance signals. The features obtained by these two convolutional neural network architectures are classified using a fully connected layer, and the power signal is processed using a 1D convolutional neural network in its original form. These signals are then converted into images and processed using a 2D convolutional neural network, combining the feature vectors generated by the 1D and 2D convolutional neural networks. Finally, this combined vector is classified by a fully connected layer. The proposed method is well suited to the nature of signal processing. This is a novel approach. The proposed framework is compared with other state-of-the-art methods for power quality disturbance classification in the literature. Although the classification performance of the proposed method is relatively high compared with other methods, the computational complexity is almost the same.