A quality control method based on physical constraints and data-driven collaborative artificial intelligence for wind observations along high-speed railway lines

Abstract

This study proposed a new quality control method via physical constraints and data-driven collaborative artificial intelligence (PD-BX) to reduce wind speed measurement errors caused by the complex environment along high-speed railway lines, achieving enhanced accuracy and reliability. On the one hand, based on the special structure in railway assembly, the physical constraint model of the railway electrical catenary supports and anemometers was experimentally established. The performance of the physical model in the wind field was simulated based on FLUENT software, and the environmental change characteristics of the anemometer in the railway area were analyzed. On the other hand, to solve the constrained error mapping expression under different wind conditions, a data-driven model of hyperparameter optimization (BO-XGBoost) is introduced to perform error compensation on physical relationships. Through the PD-BX method, the RMSE of the railway anemometer was reduced by 2.497 from 2.790 to 0.293, achieving quality control of wind observations along the high-speed railway lines and providing reliable results for improving the accuracy of the high-speed railway early warning system.