Self-Supervised Temporal Super-Resolution of Energy Data using Generative Adversarial Transformer
Xuanhao Mu, Gökhan Demirel, Yuzhe Zhang, Jianlei Liu, Thorsten Schlachter, Veit Hagenmeyer
TL;DR
This paper addresses the temporal granularity mismatch in energy system co-simulation by introducing a self-supervised Generative Adversarial Transformer (GAT) for time-series upsampling without high-resolution ground-truth data. A Transformer-based generator with a convolutional multi-head attention module collaborates with a feature-space discriminator and a multi-term loss (including feature matching) to preserve global structure and local dynamics. The method is trained in three stages and demonstrates consistent improvements over traditional baselines across multiple real-energy datasets, including an MPC application where upsampling improves control performance by significant margins. The approach offers a practical, ground-truth-free path to high-resolution energy data and has potential for real-world integration in co-simulation workflows and grid optimization.
Abstract
To bridge the temporal granularity gap in energy network design and operation based on Energy System Models, resampling of time series is required. While conventional upsampling methods are computationally efficient, they often result in significant information loss or increased noise. Advanced models such as time series generation models, Super-Resolution models and imputation models show potential, but also face fundamental challenges. The goal of time series generative models is to learn the distribution of the original data to generate high-resolution series with similar statistical characteristics. This is not entirely consistent with the definition of upsampling. Time series Super-Resolution models or imputation models can degrade the accuracy of upsampling because the input low-resolution time series are sparse and may have insufficient context. Moreover, such models usually rely on supervised learning paradigms. This presents a fundamental application paradox: their training requires the high-resolution time series that is intrinsically absent in upsampling application scenarios. To address the mentioned upsampling issue, this paper introduces a new method utilizing Generative Adversarial Transformers (GATs), which can be trained without access to any ground-truth high-resolution data. Compared with conventional interpolation methods, the introduced method can reduce the root mean square error (RMSE) of upsampling tasks by 10%, and the accuracy of a model predictive control (MPC) application scenario is improved by 13%.