<p>Tunnel fires involving multiple ignition sources pose significant challenges for accurate fire source localization due to the complexity of smoke propagation and environmental noise. This study proposes an adaptive sparrow search algorithm (SSA) for dual-source localization in tunnel fires, utilizing multi-source data fusion and information entropy-based dynamic weighting. The method integrates temperature and CO concentration data, dynamically adjusting their weights based on entropy calculations, thereby mitigating the influence of sensor noise and environmental uncertainties. Additionally, the global search mechanism of the butterfly optimization algorithm (BOA) is incorporated into the SSA, enhancing its optimization capability. The method is validated through a full-scale experiment and numerical simulation, demonstrating superior accuracy and robustness compared to traditional methods. Experimental results show that the adaptive SSA significantly reduces the localization error, and the localization results are stable and superior to other algorithms. This study provides a novel approach to dual-source localization in tunnel fires, offering high adaptability for real-time fire rescue operations.</p>