基于迁移学习和AlexNet的复合电能质量扰动识别

Compound Power Quality Disturbance Identification Based on Migration Learning and AlexNet

针对传统电能质量扰动特征提取困难的问题,提出了1种基于迁移学习与深度卷积网络相结合的复合电能质量扰动识别方法。该方法利用相空间重构法将一维扰动信号转化为形状特征明显的轨迹图像,输入到迁移学习的AlexNet模型中进行端到端的识别分类。首先,将扰动信号重构到高维相空间。然后,将其映射到二维平面以获得重构信号轨迹图像。接着,将轨迹图像输入到迁移学习的AlexNet中训练学习以实现扰动识别。最后,为了验证该方法的有效性,开展了相关仿真研究。仿真结果表明,所提方法对复合扰动的识别准确率较高。与其他网络模型相比,在保持训练参数不变的情况下,该方法具有较高的识别精度和收敛性。同时,在信号自动识别领域,该方法也提供了新的解决思路。

A method of compound power quality disturbance identification based on the combination of migration learning and deep convolutional networks is proposed to address the difficulty of traditional power quality disturbance feature extraction.The method uses phase space reconstruction to transform the one-dimensional disturbance signal into a trajectory image with distinct shape features and input it into the AlexNet model of migration learning for end-to-end recognition and classification.Firstly,the disturbance signal is reconstructed into a high-dimensional phase space.Then,the reconstructed signal trajectory image is obtained by mapping it to a two-dimensional plane.Next,the trajectory images are input to the AlexNet for migration learning to train learning to achieve disturbance recognition.Finally,to verify the effectiveness of the method,a related simulation study is carried out.The simulation results show that the proposed method has a high accuracy in recognizing composite disturbance.Compared with other network models,the method has higher recognition accuracy and convergence while keeping the training parameters unchanged.Also,the method provides a new solution idea in the field of automatic signal recognition.