Measurement and Prediction of Indoor Wideband Electric Field Radiation

Indoor wideband electromagnetic radiation (EMR) over seven days and potential radiation sources were measured and analyzed in this article. The research results show that Wi-Fi and base station equipment are the main factors causing indoor electromagnetic exposure, and the radiation intensity of the former is much higher than the latter when the equipment is working. Also, the change of EMR over time conforms to the users' daily work and rest cycle and contains a significant low-frequency component with a 24-h cycle and complex high-frequency components. Besides, the intensive use of electronic products by users has been proven to have a great impact on the frequency characteristics of the electric field, and the complexity of the time-frequency characteristics of EMR time series has increased, resulting in poor prediction results of the models such as the seasonal autoregressive moving average (SARIMA) model and the long short-term memory (LSTM) neural network. To this end, a prediction scheme combining wavelet transform and time-series analysis methods is proposed in this article. To verify the effectiveness of the proposed hybrid method, the samples of three days are exploited to perform a loop test of the prediction accuracy under different prediction steps. Besides, the results are compared with those obtained by single SARIMA and LSTM. The experimental results show that the proposed hybrid prediction method effectively predicts electric field radiation with complex frequency characteristics and significantly outperforms other methods in terms of prediction accuracy and prediction length.