在遥感海浪数据质量控制研究中,由于数据的复杂与不规则性,传统质量控制方法对海浪数据单点异常值的检测具有一定局限性。深度学习具有强大的特征学习能力,在解决非线性复杂问题方面具有一定优势,将其应用在数据质量控制领域可以提高异...在遥感海浪数据质量控制研究中,由于数据的复杂与不规则性,传统质量控制方法对海浪数据单点异常值的检测具有一定局限性。深度学习具有强大的特征学习能力,在解决非线性复杂问题方面具有一定优势,将其应用在数据质量控制领域可以提高异常值检测能力。本研究采用遥感海浪有效波高数据,构建双向长短期记忆网络(bi-directional long short term memory,Bi-LSTM)模型对有效波高进行预测,结合阈值方法进行异常检测,与3σ准则法、孤立森林模型法、 LSTM模型法以及VAE-LSTM模型法进行异常检测精度比较,探究基于Bi-LSTM的质量控制模型在遥感海浪数据异常值检测方面的能力。试验结果表明,Bi-LSTM质量控制模型具有良好的异常值检测效果,其精准率、召回率、 F1分数和运行时间分别为91%、 93%、 92和3.35 s,综合评价效果最佳,可有效对遥感海浪数据进行质量控制。展开更多
光伏发电功率的准确预测对于优化能源管理和电网规划及优化调度具有重要的意义。针对以往光伏发电功率预测方法预测精度不高,传统混合网络模型存在参数选择不确定性和收敛速度较慢的问题,基于历史气象数据和光伏发电数据,提出一种结合...光伏发电功率的准确预测对于优化能源管理和电网规划及优化调度具有重要的意义。针对以往光伏发电功率预测方法预测精度不高,传统混合网络模型存在参数选择不确定性和收敛速度较慢的问题,基于历史气象数据和光伏发电数据,提出一种结合向量加权平均(weighted mean of vectors,INFO)算法、卷积网络(convolutional neural network,CNN)和双向长短期记忆网络(bi-directional long short-term memory,BiLSTM)的光伏发电功率预测方法。首先,选取与光伏发电功率预测相关的多种气象因素,含太阳辐射、温度、湿度、风速、气压等气象参数,并分析它们与光伏发电功率之间的关系,然后使用INFO算法对CNNBiLSTM混合网络预测模型的隐藏层节点数、初始学习率和L2正则化系数进行优化,INFO算法通过自适应调整这些参数,缩短了手动调制参数的时间,提高了超参数设置的精度和效率。实验结果表明,通过INFO算法优化的CNN-BiLSTM混合网络相比传统CNN-BiLSTM混合网络具有更高的预测精度。展开更多
Interplay between topology and magnetism can give rise to exotic properties in topological materials.Two-dimensional bismuth has been extensively studied owing to its topological states with a strong spin-orbit coupli...Interplay between topology and magnetism can give rise to exotic properties in topological materials.Two-dimensional bismuth has been extensively studied owing to its topological states with a strong spin-orbit coupling,and 1T-VTe_(2)monolayer theoretically predicted to host an intrinsic magnetism as experimentally suggested.In this work,we successfully constructed a vertical heterostructure composed of the two-dimensional Bi(110)monolayer and 1T-VTe_(2)monolayer by using molecular beam epitaxy(MBE).Scanning tunneling microscopy(STM)measurements revealed that the growth of Bi preferably occurs along the step edges of the VTe_(2)monolayer,forming a Bi(110)monolayer on top of the VTe_(2)monolayer next to a peripheral Bi bilayer.The Bi(100)/VTe_(2)heterostructure exhibits a specific lattice registry with a well-defined moiréperiodicity.Scanning tunneling spectroscopy(STS)measurements further unveiled an universal suppression in the local density-of-states at the boundary of the Bi(110)/VTe_(2)bilayer.By examining the atomic structures of Bi(110)boundaries,we found this effect does not originate from the previously proposed atomic reconstruction at the step edge of Bi(110),but is likely related to the magnetic properties of the VTe_(2)monolayer.展开更多
文摘在遥感海浪数据质量控制研究中,由于数据的复杂与不规则性,传统质量控制方法对海浪数据单点异常值的检测具有一定局限性。深度学习具有强大的特征学习能力,在解决非线性复杂问题方面具有一定优势,将其应用在数据质量控制领域可以提高异常值检测能力。本研究采用遥感海浪有效波高数据,构建双向长短期记忆网络(bi-directional long short term memory,Bi-LSTM)模型对有效波高进行预测,结合阈值方法进行异常检测,与3σ准则法、孤立森林模型法、 LSTM模型法以及VAE-LSTM模型法进行异常检测精度比较,探究基于Bi-LSTM的质量控制模型在遥感海浪数据异常值检测方面的能力。试验结果表明,Bi-LSTM质量控制模型具有良好的异常值检测效果,其精准率、召回率、 F1分数和运行时间分别为91%、 93%、 92和3.35 s,综合评价效果最佳,可有效对遥感海浪数据进行质量控制。
文摘光伏发电功率的准确预测对于优化能源管理和电网规划及优化调度具有重要的意义。针对以往光伏发电功率预测方法预测精度不高,传统混合网络模型存在参数选择不确定性和收敛速度较慢的问题,基于历史气象数据和光伏发电数据,提出一种结合向量加权平均(weighted mean of vectors,INFO)算法、卷积网络(convolutional neural network,CNN)和双向长短期记忆网络(bi-directional long short-term memory,BiLSTM)的光伏发电功率预测方法。首先,选取与光伏发电功率预测相关的多种气象因素,含太阳辐射、温度、湿度、风速、气压等气象参数,并分析它们与光伏发电功率之间的关系,然后使用INFO算法对CNNBiLSTM混合网络预测模型的隐藏层节点数、初始学习率和L2正则化系数进行优化,INFO算法通过自适应调整这些参数,缩短了手动调制参数的时间,提高了超参数设置的精度和效率。实验结果表明,通过INFO算法优化的CNN-BiLSTM混合网络相比传统CNN-BiLSTM混合网络具有更高的预测精度。
基金financially supported by the National Key Research and Development Program of China(Grant No.2021YFA1400403)the National Natural Science Foundation of China(Grant Nos.12374183,92165205)+2 种基金the Natural Science Foundation of Jiangsu Province(Grant No.BK20233001)the Innovation Program for Quantum Science and Technology(Grant No.2021ZD0302800)the Fundamental Research Funds for the Central Universities(Grant No.020414380207).
文摘Interplay between topology and magnetism can give rise to exotic properties in topological materials.Two-dimensional bismuth has been extensively studied owing to its topological states with a strong spin-orbit coupling,and 1T-VTe_(2)monolayer theoretically predicted to host an intrinsic magnetism as experimentally suggested.In this work,we successfully constructed a vertical heterostructure composed of the two-dimensional Bi(110)monolayer and 1T-VTe_(2)monolayer by using molecular beam epitaxy(MBE).Scanning tunneling microscopy(STM)measurements revealed that the growth of Bi preferably occurs along the step edges of the VTe_(2)monolayer,forming a Bi(110)monolayer on top of the VTe_(2)monolayer next to a peripheral Bi bilayer.The Bi(100)/VTe_(2)heterostructure exhibits a specific lattice registry with a well-defined moiréperiodicity.Scanning tunneling spectroscopy(STS)measurements further unveiled an universal suppression in the local density-of-states at the boundary of the Bi(110)/VTe_(2)bilayer.By examining the atomic structures of Bi(110)boundaries,we found this effect does not originate from the previously proposed atomic reconstruction at the step edge of Bi(110),but is likely related to the magnetic properties of the VTe_(2)monolayer.
