Against the backdrop of synergistic development between intelligent industrial processes and immersive interactive technologies, auditory perceptual load in high-concurrency multi-source environments has become a bottleneck affecting the efficacy of virtual simulation training systems. To address the limited static adaptability and cross-modal latency of existing sound source localization methods in dynamic scenarios, this study proposes an optimization algorithm based on a psychoacoustic dynamic perception model. By establishing a nonlinear correlation framework between sound source density and perceptual thresholds, the algorithm transcends the perceptual boundary constraints of traditional geometric calibration methods and innovatively designs a Resource Competition Attenuation Factor to dynamically allocate spatial auditory attention. Experimental validation demonstrates that this method significantly enhances azimuth identification stability and operational response efficiency in complex sound fields, providing a multi-source perception enhancement solution for industrial virtual training systems that balances real-time performance with robustness.