基于凌日搜索优化CNN/BI-GRU的电能质量扰动分类方法  

Power Quality Disturbance Classification Method Based on Transit Search Optimization CNN/BI-GRU

作　　者：高帅 杨永超 童占北 钟建伟 GAO Shuai;YANG Yongchao;TONG Zhanbei;ZHONG Jianwei(College of Intelligent Systems Science and Engineering,Hubei Minzu University,Enshi 445000,China)

机构地区：[1]湖北民族大学智能科学与工程学院,湖北恩施445000

出　　处：《湖北民族大学学报（自然科学版）》2024年第3期361-367,共7页Journal of Hubei Minzu University:Natural Science Edition

基　　金：湖北省自然科学基金项目(2022CFB264);恩施州技术支撑类科技计划项目(D20230042)。

摘　　要：针对复杂电能质量扰动分类方法识别准确率不高的问题,提出了一种基于凌日搜索优化多模态网络模型的电能质量扰动分类方法。首先,利用格拉姆角场对初始一维时序信号进行数据处理得到二维图像数据;然后,分别将时序信号与图像数据输入多模态网络中进行特征提取,利用凌日搜索算法优化多模态网络参数,提升网络特征捕获能力;再通过特征融合模块,将时序特征和图像特征有效融合;最后,利用自注意力机制增强网络模型对下文信息的理解能力。结果表明,在无噪声环境下分类准确率达到99.2%,在不同信噪比环境下平均分类准确率达到98.3%。该研究能对新型电力系统中愈加复杂的电能质量扰动实现准确的分类,与传统分类方法相比鲁棒性较强。Aiming at the problem of low identification accuracy of complex power quality disturbance classification methods,a method for power quality identification and classification based on the transit search optimized multi-modal network model was proposed.Firstly,the Gramian angular field was used to perform data processing on the initial one-dimensional time series signal to obtain two-dimensional image data.Secondly,the time series signal and image data were input into the multi-modal network for feature extraction,and the transit search algorithm was used to optimize the parameters of the multimodal network to improve the feature capture capability of the network.Then,through the feature fusion module,the time series features and image features were fused effectively.Finally,the self-attention mechanism was used to enhance the network model′s ability to understand contextual information.The results showed that the method proposed in this paper had a classification accuracy of 99.2%in a noise-free environment,and an average classification accuracy of 98.3%in different signal-to-noise ratio environments.The proposed method achieves accurate classification of increasingly complex power quality disturbances in new power systems,and it is more robust than traditional classification methods.

关 键 词：电能质量扰动 深度学习 格拉姆角场 特征融合 凌日搜索算法 自注意力机制 

分 类 号：TM711[电气工程—电力系统及自动化]