High-fidelity fiber longitudinal power monitoring via non-uniform sparse regularization

Sparse regularization algorithms for longitudinal power monitoring (LPM) often exhibit distortion in low-power regions and excessive smoothing at power discontinuities. To overcome these limitations, we propose a high-fidelity monitoring method based on physics-aware non-uniform sparse regularization. By incorporating asymmetric constraints and a spatial point decoupling strategy, the proposed method effectively suppresses non-physical power rise in low-power regions and preserves sharp transitions at the erbium-doped optical fiber amplifier (EDFA). With few measurements, the proposed method achieves a 0.58-dB reduction in root mean square error (RMSE) compared to the two-stage sparse regularization method and 1 km localization accuracy, significantly improving the fidelity and robustness of LPM.