To enhance the capabilities of the Inner Mongolia Earthquake Early Warning System in data quality evaluation and frequency-domain anomaly detection, this study employs the Power Spectral Density (PSD) values corresponding to the 10th and 90th percentiles of the background noise power spectral density Probability Density Function (PDF) as a method to establish station background noise models. Utilizing the full year 2024 of continuous observation data from 48 upgraded reference stations in Inner Mongolia, we have quantitatively derived the background noise models (L-PSD and H-PSD) for these 48 stations for the first time. Based on these individual station models, we further established regional background noise models for Inner Mongolia (MODEL-PSD-1 and MODEL-PSD-2). These models are intended to support station operation monitoring, data anomaly identification, and scientific site selection. The findings demonstrate that: (1) Compared to global noise models (NHNM/NLNM), the locally developed station-level noise models exhibit higher sensitivity in data quality assessment, effectively identifying various anomalies, including pendulum cable faults in seismometers, changes in installation configurations, environmental interference, and differences in frequency response; (2) The established regional noise models show strong regional representativeness and seasonal adaptability, encompassing noise variation characteristics across diverse observational environments. They provide a reliable quantitative regional benchmark for site selection, observational environment assessment, and data quality inspection of new stations in Inner Mongolia and adjacent areas; (3) Noise source analysis indicates that the low-frequency band (0.02~0.05 Hz) is predominantly influenced by airflow and temperature variations, the microtremor band (0.1~0.5 Hz) is closely associated with oceanic activity, and the high-frequency band (1~20 Hz) is mainly affected by anthropogenic sources. The noise models developed in this study significantly improve the capacity for data quality evaluation and frequency-domain anomaly detection within the seismic network of Inner Mongolia and surrounding regions. They also establish a scientific foundation for future deployment of intelligent quality monitoring systems and the selection of sites for new stations.