安徽大学学报（自然科学版）2025，Vol.49Issue(4)：57-65,9.DOI:10.3969/j.issn.1000-2162.2025.04.007
基于多头卷积注意力网络的电能质量扰动识别算法
Power quality disturbance identification method based on multi-head attention convolutional network
摘要
Abstract
The diverse types of power quality disturbance signals and strong feature coupling bring enormous challenges to power quality disturbance identification and classification.To improve the recognition accuracy of power quality disturbance signals in complex environments,a power quality disturbance recognition algorithm based on a multi-head convolutional attention network was proposed.First,to reduce the complexity of the model,a multi-scale residual convolution module was designed to reconstruct the disturbance signal features.Then,to address issues such as unclear recognition caused by the limited sensory field of the convolution kernel,a multi-head convolutional attention module was designed to capture the global dependency of the disturbance signal time series and achieve the fusion of the global features of the disturbance signal with local features of different scales.Finally,a fully connected layer was used to map the high-dimensional feature information to low-dimensional space,and the Softmax classifier was applied to identify and classify the power quality disturbance signals.The experimental results showed that the proposed algorithm extracted the features of single and composite power quality disturbance signals effectively and demonstrated higher accuracy under different SNRs compared with other algorithms,verifying the robustness of the algorithm.关键词
电能质量扰动/卷积神经网络/Transformer网络/注意力机制Key words
power quality disturbance/convolutional neural network/transformer network/attention mechanism分类
信息技术与安全科学引用本文复制引用
安林,郑玉平,竺德,李明胜,高清维..基于多头卷积注意力网络的电能质量扰动识别算法[J].安徽大学学报（自然科学版）,2025,49(4):57-65,9.基金项目
电网运行风险防御技术与装备全国重点实验室科研项目(SGNR0000KJJS2302144) （SGNR0000KJJS2302144）
