Weather-Adaptive Multi-Step Forecasting of State of Polarization Changes in Aerial Fibers Using Wavelet Neural Networks

TL;DR

This paper presents a weather-adaptive wavelet neural network method for multi-step forecasting of polarization changes in aerial fiber links, significantly improving prediction accuracy by integrating weather data and multi-scale analysis.

Contribution

The novel integration of weather data with wavelet neural networks for enhanced multi-step SOP forecasting in aerial fibers.

Findings

Over 65% RMSE improvement

Over 63% MAPE improvement

Effective multi-scale polarization prediction

Abstract

We introduce a novel weather-adaptive approach for multi-step forecasting of multi-scale SOP changes in aerial fiber links. By harnessing the discrete wavelet transform and incorporating weather data, our approach improves forecasting accuracy by over 65% in RMSE and 63% in MAPE compared to baselines.