Long-term Time Series Prediction of Deformation in The Area of Pylons By Combining InSAR and Transformer-based

Settlement monitoring of high-voltage line pylons (HLPs) is an important part of power line inspection. In geologically active areas, or due to some environmental factors, it can substantially impact the stability of HLP bases, which can lead to collapse accidents, so it is essential to monitor and predict the areas around HLPs. In this paper, we use 74-view Sentinel-1A satellite ascending orbit data, and combine the persistent scatterer interferometric synthetic aperture radar (PS-InSAR) with a long-term sequence prediction model (Autoformer) to perform surface deformation of a high-voltage line in Inner Mongolia, China, from January 2019 to January 2022 monitoring. The environmental factors were considered to predict the surface deformation in the key study area from February 2022 to April 2022 and provide early warning. By comparing with Transformer, Informer, and Reformer, the results show that the proposed combined PS-InSAR and Autoformer method is far superior to its counterparts with a mean squared error of 0.105. This paper provides a new, accurate, and automated method for HLP subsidence monitoring and early warning and establishes a system for HLPs surface deformation prediction, which highly improves the efficiency of HLP monitoring and early warning.