To address the limitations of traditional image retrieval methods that are predominantly constrained to single-phase data and lack comprehensive research on time-series remote sensing images, a novel change information retrieval model, SCanNet-Retrieval( Semantic Change Network and Retrieval) is proposed, which aims to enhance the performance of change information retrieval for dual-phase images. The architecture of SCanNet-Retrieval comprises two primary modules, the feature extraction module and the similarity measurement module. The feature extraction module integrates an encoder-decoder structure with the SCanFormer module and incorporates a category change matrix to effectively capture spatiotemporal semantic change features. In the similarity measurement module, the Jaccard similarity coefficient is employed to assess retrieval performance. Three other similarity measurement methods, Euclidean distance, Manhattan distance, and Hamming distance are compared with validate the effectiveness of the proposed model. To address the scarcity of publicly available two-phase datasets in the domain of energy infrastructure, the EICIRD(Energy Infrastructure Change InformationRetrieval Dataset) is constructed. Experimental results indicate that SCanNet-Retrieval achieves an average retrieval accuracy exceeding 93% across all change categories, significantly outperforming other methods. This underscores its potential for efficient and accurate retrieval of energy infrastructure change information from large-scale time-series image data. This method offers critical support for the intelligent monitoring of energy infrastructure and the green transformation of the energy industry.