基于WD-LSTM的宽带电磁辐射时序建模预测方法  

Method for wideband electromagnetic radiation time series modeling prediction based on WD-LSTM

作　　者：杨晨 宋欣蔚 岳云涛[1] YANG Chen;SONG Xinwei;YUE Yuntao(School of Intelligence Science and Technology,Beijing University of Civil Engineering and Architecture,Beijing 100044,China)

机构地区：[1]北京建筑大学智能科学与技术学院,北京100044

出　　处：《现代电子技术》2025年第6期9-15,共7页Modern Electronics Technique

摘　　要：无线通信技术的飞速发展以及包含相关功能产品的广泛使用,使得环境电磁场呈现复杂的变化特性,且城市电磁环境状况日益恶化,故进行电磁辐射的分析与预测对于潜在风险预警与控制至关重要。文中对北京市典型商业区核心街道连续时段的宽带电磁辐射进行了测量,并对其进行了短时傅里叶变换分析。分析结果显示,电磁辐射时变规律与人们的作息活动具有相关性,且受部分时段无线设备密集使用的影响,呈现出强烈的低频周期性和高频波动性,而这些特性会导致单一的时序建模方法预测效果变差。为此,提出了一种结合小波分解(WD)与长短时记忆(LSTM)模型的混合预测方法。该方法根据电磁辐射时频特性,将其分解为主要周期分量和细节分量进行分层预测,以适应复杂城市电磁环境状况。基于测量数据,将所提方法与其他典型时序预测模型进行对比,结果表明,该方法的预测准确度更高,并具有更强的异常值适应性与稳定性。The rapid development of wireless communication technology and the wide use of products containing related functions make the environmental electromagnetic field present complex changing characteristics,and the urban electromagnetic environment is deteriorating day by day.Therefore,the analysis and prediction of electromagnetic radiation is of great importance for potential risk warning and control.The broadband electromagnetic radiation in the core street of typical business district of Beijing is measured and analyzed by means of the short time Fourier transform(STFT).The analysis results show that the timevarying rule of electromagnetic radiation is correlated with people′s rest and rest activities,and it presents a strong low-frequency periodicity and high-frequency volatility due to the intensive use of wireless devices in some periods,which will lead to the poor prediction effect of a single time-series modeling method.On this basis,a hybrid prediction method combining Wavelet decomposition(WD)and long short-term memory(LSTM)model is proposed.The method is based on the time-frequency characteristics of electromagnetic radiation,which is decomposed into main period components and detail components for the hierarchical prediction to adapt to the complex urban electromagnetic environment conditions.The proposed method is compared with other typical time-series prediction models based on measured data.The results show that the proposed method has higher prediction accuracy and stronger outlier adaptability and stability.

关 键 词：宽带电磁辐射 时间序列 小波分解 长短时记忆模型 时频特性 分层预测 

分 类 号：TN98-34[电子电信—信息与通信工程]