Study on field strength prediction using different models on time series from urban continuous RF- EMF monitoring

Authors: Song X, Feng W, Yang C, Djuric N, Kljajic D, Djuric S

Year: 2025

Category: Environmental Health, Machine Learning, Electromagnetic Field Safety

Journal: Expert Systems with Applications

Institution: Not specified

DOI: 10.1016/j.eswa.2025.126963

URL: https://www.sciencedirect.com/science/article/pii/S0957417425005858

Abstract

Overview

State-of-the-art electromagnetic field (EMF) monitoring networks, such as the Serbian EMF RATEL system, now enable continuous, daily monitoring of radio-frequency (RF) EMF levels. This is vital in urban areas where prolonged human presence can lead to heightened sensitivity to RF-EMF exposure.

Value of Near-Future RF-EMF Prediction

  • Provides critical support for public health initiatives.
  • Supplements EMF monitoring in locations identified as high-risk.
  • Assists in proactively reducing the duration of human exposure to high EMF levels.
  • Facilitates advanced testing for EMF compliance, focusing on the areas with vulnerable populations.

Findings

  • This paper evaluates several models for predicting field strength in urban environments: SARIMA, CNN, LSTM, ELM, PLS Regression, and Transformer networks.
  • The models are tested on time series data from two kindergartens and one elementary school in Novi Sad, Serbia, emphasizing the need to monitor environments inhabited by sensitive populations.
  • Analysis is based on two years of continuous EMF monitoring data.
  • Metrics evaluated include prediction accuracy, performance degradation rate, accuracy in extreme value prediction, and model training time.
  • The Partial Least Squares Regression (PLS) model shows superior performance in predicting EMF exposure compared to others.

Conclusion

  • This preliminary analysis offers a valuable framework for large-scale real-time EMF monitoring in urban environments.
  • It highlights the significant link between urban RF-EMF exposure and public health risk, justifying the need for ongoing monitoring and predictive research.
  • The study serves as a foundation for future work aimed at protecting public health from electromagnetic field exposure risks.
RF-EMF electromagnetic field urban monitoring machine learning models public health time series prediction exposure risk
← Back to Stats