基于递归图深度学习的电能质量异常扰动检测

Detection of Abnormal Disturbances in Power Quality Based on Recursive Graph Deep Learning

电能质量问题会严重影响供电质量。为解决该问题,本文研究了电能质量的识别。首先,提出了一种基于数据预处理递归图的特征提取方法,用于提取各类电能质量异常扰动的特征;然后,提出了一种基于机器学习的电能质量异常扰动检测方法,利用Swin-Transformer网络高效表征递归图中的隐含信息,同时降低模型计算的复杂度,增加对电能质量事件的识别率和准确率;最后,通过MATLAB-Simulink仿真模拟电能质量扰动数据并进行实验,验证了该方法的有效性。结果表明,数据预处理递归图能有效提高不同电能质量异常扰动类型的特征区分度。Swin-Transformer网络能高效且精确地识别电能质量异常扰动。

Power quality issues can seriously affect the quality of power supply.To solve the problem,this paper investigates the identification of power quality.Firstly,a feature extraction method based on data preprocessing recursive graph was proposed to extract features of various types of abnormal disturbances in power quality.Then,a machine learning based method for detecting abnormal disturbances in power quality was proposed,which utilizes the Swin-Transformer network to efficiently represent the implicit information in the recursive graph,while reducing the complexity of model calculations and increasing the recognition and accuracy of power quality events.Finally,the effectiveness of this method was verified through MATLAB Simulink simulation of power quality disturbance data and experiments.The results show that data preprocessing recursive graphs can effectively improve the feature discrimination of different types of abnormal disturbances in power quality.The Swin-Transformer network can efficiently and accurately identify abnormal disturbances in power quality.