AUTHOR=Xu Jinxing , Ji Zhenya , Liu Xiaofeng , Bao Yuqing , Zhang Shiwei , Wang Wei , Pang Zihao TITLE=Two-stage scheduling of integrated energy systems based on a two-step DCGAN-based scenario prediction approach JOURNAL=Frontiers in Energy Research VOLUME=Volume 10 - 2022 YEAR=2023 URL=https://www.frontiersin.org/journals/energy-research/articles/10.3389/fenrg.2022.1012367 DOI=10.3389/fenrg.2022.1012367 ISSN=2296-598X ABSTRACT=Integrated energy systems (IESs) are developing rapidly as a supporting technology for achieving carbon reduction targets. IESs involve a large degree of uncertainties of sources and loads. An accurate prediction can facilitate better to scheduling strategies. Recently, a new-developed unsupervised machine learning tool, i.e., Generative Adversarial Networks (GAN), has been used to predict renewable energy outputs and various types of loads. GAN has an advantage in predicting sources and loads of IES due to the fact that no prior assumptions about the data distribution are required. However, on the one hand, the structure of the traditional GAN leads to the problem of uncontrollable generations, which needs to be improved. On the other hand, the generated scenarios bring with it a larger amount of computation that needs to be retained. Therefore, we propose a two-step prediction approach that takes deep convolutional GAN (DCGAN) to achieve higher accuracy generation results and uses a K-means clustering algorithm to achieve scenario reduction. Common two-stage scheduling is generally day-ahead and intraday stages, with rolling scheduling used for the intraday stage. In this paper, the predictions of DCGAN are used as the sources and loads of the inputs. To account for the impacts of prediction accuracy of scheduling results, Conditional Value at Risk (CVaR) is added to the day-ahead stage. The process of the intra-day prediction has also been improved to ensure that the inputs are updated in real time in each prediction domain. Simulations on a typical IES show that the proposed method can improve the accuracy and economy of IES scheduling results.