This paper introduces the Stream Query Denoising (SQD) strategy to enhance temporal modeling in the construction of vectorized high-definition (HD) maps for autonomous driving. The SQD strategy is designed to improve the learning of temporal consistency among map elements within streaming models, particularly in the context of HD map construction. The methodology involves denoising queries that have been perturbed by adding noise to the ground-truth information from the preceding frame, aiming to reconstruct the ground-truth information for the current frame. This process simulates the prediction process inherent in stream queries. The proposed SQD-MapNet, which integrates SQD into the StreamMapNet framework, demonstrates superior performance over existing methods on both nuScenes and Argoverse2 datasets across various settings of close and long ranges. The paper also includes extensive ablation studies and qualitative comparisons to validate the effectiveness of the proposed approach.This paper introduces the Stream Query Denoising (SQD) strategy to enhance temporal modeling in the construction of vectorized high-definition (HD) maps for autonomous driving. The SQD strategy is designed to improve the learning of temporal consistency among map elements within streaming models, particularly in the context of HD map construction. The methodology involves denoising queries that have been perturbed by adding noise to the ground-truth information from the preceding frame, aiming to reconstruct the ground-truth information for the current frame. This process simulates the prediction process inherent in stream queries. The proposed SQD-MapNet, which integrates SQD into the StreamMapNet framework, demonstrates superior performance over existing methods on both nuScenes and Argoverse2 datasets across various settings of close and long ranges. The paper also includes extensive ablation studies and qualitative comparisons to validate the effectiveness of the proposed approach.