Research on Coarse Granularity Data Sample Completion Method for District Heating System

Abstract

The heating system is an essential part of China’s urban energy system. However, the heating systems in China have varying levels of hardware and software, informationization, and intelligence. The coarse time granularity of actual system data and the missing data brings significant challenges to system analysis and diagnosis. This paper proposes to study the sample completion and granularity refinement of the time-series data of heating systems based on Wasserstein Generative Adversarial Networks (WGAN). The method is validated based on the field data from 2019 to 2020 of a district heating system in Zhengzhou. Besides, the key features affecting the quality of the completion are discussed. After introducing the data processing method of Empirical Modal Decomposition (EMD), the results show that historical data and climate data dominate the quality of the generated samples. The WGAN can generate the samples accurately. When comparing the real data under the same conditions, the accuracy rate reaches 96.4%. This study further refines the sampling granularity of the existing heating system based on the WGAN to provide more accurate samples for system analysis and diagnosis. Overall, the study provides a new data analysis method for theoretical and technical studies under data deficiency scenarios for heating systems.