An ELM data-driven model for predicting erosion rate of string in underground compressed air storage

Abstract

Underground compressed air storage has experienced rapid development in recent years. The string, as a critical component of underground compressed air storage, serves as a channel for gas injection and production. The high-speed gas flow can cause string failure due to erosion. This study presents a data-driven model to predict the erosion rate of string in underground compressed air storage. A Computational Fluid Dynamics (CFD) model coupling discrete phase and continuous phase is developed to estimate the erosion rate of string. The effect of gas velocity, bending angle of string, string diameter, sand diameter and sand mass flow rate is investigated by implementing numerous simulations to generate a datasets of string erosion rate and its influencing factors. This dataset is used to develop a data-driven model based on extreme learning machine (ELM) algorithm for rapidly predicting the erosion rate of string. The developed model exhibits advantages in robustness and prediction accuracy.