T2SR: Super-Resolution in Smart Meter Data Using a Transformer-Based Framework

Abstract

Access to smart metre data at higher resolutions has the potential of improving energy management and load forecasting. However, such data presents several complexities, such as added pressure on resources and increased expenses. Super-Resolution (SR) is a technology with the capability of solving this problem, by reconstructing low-resolution data into high-resolution data. This study examines the potential of predicting high-resolution smart metre data obtained from low-resolution smart metre data. The study is conducted in Bergen, Norway, where power consumption data was used from a residential building. To tackle the challenges of acquiring high-resolution data, the authors propose a Transformer-based T2SR framework for SR in smart metre data. The proposed framework shows promising results in experiments, especially in predicting complex details in target patterns, in comparison to other state-of-the-art deep learning methods. The code is available at https://github.com/Ci2Lab/Malin_T2SR_Framework.