摘要
由于微电网数据的多样性和复杂性,导致数据特征挖掘难度大,降低了实时检索的精度,为此,提出了一种基于映射关系的多模态微电网数据实时检索方法。利用Borderline-SMOTE算法扩充微电网数据类别中少数类的样本数量。采用循环注意力机制映射关系法对扩充后的图像信息和文本信息进行对齐,使数据检索模型更容易挖掘数据特征。使用深度全连接网络构建数据实时检索网络模型,通过对齐图文信息进行分类输出检索结果,并通过softmax函数约束检索结果,使数据检索模型输出结果与实际需求一致,提升多模态微电网数据实时检索精度,以便及时发现微电网运行故障位置。实验结果表明,基于映射关系的数据检索方法能够准确找到与实际需求相匹配的数据,并且表现出强大的数据处理能力。
Due to the diversity and complexity of microgrid data,it is difficult to mine data features,which re-duces the accuracy of real-time retrieval.In this regard,a real-time retrieval method of multi-mode microgrid data based on mapping relationship is proposed.The Borderline-SMOTE algorithm is used to expand the number of sam-ples of minority classes in the microgrid data category.The extended image information and text information are a-ligned using the circular attention mechanism mapping relationship method,which makes the data retrieval model eas-ier to mine data characteristics.The deep full connection network is used to build a real-time data retrieval network model,and the retrieval results are classified and output by aligning the image and text information.The retrieval results are constrained by the softmax function,so that the output results of the data retrieval model are consistent with the actual needs,and the real-time retrieval accuracy of multi-mode micro grid data is improved,so as to find the fault location of micro grid operation in time.The experimental results show that the data retrieval method based on mapping relationship can accurately find the data matching the actual needs,and shows strong data processing ability.
作者
王璠
汪江
温炜
李月华
WANG Fan;WANG Jiang;WEN Wei;LI Yue-hua(State Grid Ningxia Electric Power Party School(State Grid Ningxia Training Center),Yinchuan Ningxia 750000,China)
出处
《计算机仿真》
2024年第11期173-177,共5页
Computer Simulation
基金
国网宁夏电力有限公司科技项目(SGTYHT/21-JS-226)。
关键词
双向编码器
微电网数据
跨模态注意单元
深度全连接网络
Borderline-SMOTE algorithm
Bidirectional GRU encoder
Microgrid data
Transmodal attention ele-ment
Deep full connection network
