Review and Outlook on Deep Learning for High-Resolution Power Time Series Reconstruction

High-resolution time series reconstruction is crucial for enhancing the utility of sparse or low-quality data, with broad applications in industrial monitoring, signal processing, and energy management. Advances in deep learning and physical modeling have propelled computer science and electrical engineering to develop diverse reconstruction techniques, yet their progress and paradigmatic differences remain unsystematically compared. This review contrasts these fields through methodological innovations and application scenarios, analyzing their technical characteristics and integration trends. By investigating the current state of interdisciplinary research, this study offers methodological guidance and practical insights for advancing high-resolution reconstruction of power time series, focusing on photovoltaic and grid-related data, with emphasis on improving computational efficiency and real-time applicability.