Time series modeling and prediction of wideband electric field radiation

Abstract

Exposure to wideband electromagnetic radiation (EMR) over seven days was measured in this paper, and a short-time Fourier transform analysis on this showed that the change of EMR over time conforms to people’s daily work and rest cycle and is affected by the intensive use of wireless communication devices, thus showing strong low-frequency periodicity and high-frequency fluctuations. These characteristics lead to a significant decrease in the performance of traditional EMR time series mod-els. To this end, a hybrid model based on wavelet decomposition (WD) and long-short time memory (LSTM) is proposed for the wideband electric field radiation prediction. The proposed model is verified by the measurement samples under different prediction steps. It is also compared with other representative models, and the results show that the proposed model outperforms other models in prediction accuracy and prediction steps.