基于时空知识关联性深度挖掘的频谱能量预测方法研究

Spectrum energy prediction method based on deep mining of temporal and spatial knowledge correlation

频谱能量预测是实现有限频谱资源高效利用的重要途径,而对历史电磁频谱数据进行多尺度、多维度的时空知识提取和关联关系挖掘,从而形成电磁环境画像是实现精准频谱能量预测的重要基础。对于单采集点的时间动态频谱能量数据,时域知识关联关系挖掘受到频谱数据体量大及特征维度高的影响,本文提出了基于并行多模型融合的单点时域特征提取和预测方法;对于多点采集形成的区域频谱能量数据,受到时间动态和空间分布不均的双重影响造成时空知识关联关系挖掘困难,本文首先基于区域电磁环境的相关性构建区域电磁环境的相关关系图,然后基于关系图的关联信息设计了基于图卷积的网络预测模型,研究电磁环境中频谱能量预测问题。通过仿真实验,验证了无论是在单点时间动态场景下还是区域时空场景下,本文提出的方法均优于基线模型,具有良好的预测精度和鲁棒性。

Spectrum energy prediction is an important way to realize efficient utilization of limited spectrum resources.Multi-scale and multi-dimensional spatial-temporal knowledge extraction and correlation mining of historical electromagnetic spectrum data,so as to form an electromagnetic environment portrait,is an important basis for accurate spectral energy prediction.For the time-dynamic spectral energy data of a single collection point,the time-domain knowledge association mining is affected by the large volume of spectral data and the high feature dimension.Therefore,this paper proposes a single-point time-domain knowledge association mining method based on multi-model fusion.For the regional spectral energy data formed by multi-point collection,it is difficult to mine the spatial-temporal knowledge correlation due to the dual influence of temporal dynamics and uneven spatial distribution.Therefore,this paper constructs the correlation diagram of regional electromagnetic environment based on the correlation of regional electromagnetic environment.A network prediction model based on graph convolution is designed based on the relational information of the graph to study the spectrum energy prediction problem in electromagnetic environment.Through simulation experiments,it is verified that the proposed method is superior to the baseline model in both single-point time dynamic scenarios and regional spatiotemporal scenarios,and has good prediction accuracy and robustness.