To accurately characterize the wind field at a bridge site, structural health monitoring systems equipped with anemometers installed on the bridge's main girder and pylons conducts long-term monitoring of the wind environment, collecting a vast amount of measured data to support wind load analysis. This study, using the Runyang Bridge as an example, presents a comparative analysis of real-time wind speed and direction data recorded by anemometers located upstream and downstream at midspan. Considering the complex aerodynamic interactions between the bridge deck and the wind, the disturbance effect of the bridge deck on the measured wind data is proposed. Subsequently, disturbed wind parameters are extracted from the measured data, and the abnormal distribution patterns of these parameters are studied. To obtain the true wind parameters, this study develops a data cleaning method for mean wind speed and turbulence intensity, proposing a statistical model of disturbance component decomposition. A nonlinear optimization algorithm is employed to solve the model, yielding empirical formulas between the disturbed wind parameters and the true wind parameters. Validation results from the observed data indicate that the proposed data cleansing method effectively reduces the impact of the bridge deck's disturbance on the measured data and wind parameters, ensuring accurate evaluation of the wind loads borne by the structure.