Abstract:The geoelectric field is one of the important aspects of earthquake prediction research, and its data quality directly affects the accuracy of pre-seismic anomaly detection. With the expansion of the national rail transit networks, stray currents generated by metro operations have become one of the main sources of interference affecting geoelectric field observations. To address the issue of metro interference suppression, this study systematically compared the denoising effects of three filtering methods: moving average filtering, empirical mode decomposition (EMD) filtering, and wavelet filtering. By constructing simulated signals that closely resembled real-world scenarios, the filtering performance was evaluated using indicators such as signal-to-noise ratio (SNR) and root mean square error (RMSE). Simulation experiments showed that the moving average filtering achieved the best performance with a window length of 390 s, improving SNR by 15.50 dB and reducing RMSE by 83.22%. For EMD filtering, removing the first two intrinsic mode functions (IMFs) was more reasonable to balance denoising performance and valid signal preservation, improving SNR by 10.99 dB and reducing RMSE by 71.8%. Wavelet filtering demonstrated the best filtering performance after an 8-level decomposition using the db4 wavelet basis, improving SNR by 16.86 dB and reducing RMSE by 85.7%. Filtering processing was applied to the second-sampled geoelectric field data from the Shanghai Qingpu station affected by metro interference. The results showed that all three methods reduced the signal standard deviation by over 65%. Among them, wavelet filtering showed significant advantages in preserving the details of abrupt signal changes. The study indicates that filtering methods with optimized parameters using second-sampled data can effectively suppress rail transit interference, providing technical support for improving the quality of geoelectric field observation data.
