Daily natural gas load forecasting based on sequence autocorrelation

• Published in: 2022 37th Youth Academic Annual Conference of Chinese Association of Automation (YAC) pp. 1452 - 1459
• Main Authors: Xiang, Xingren, Shen, Jiayuan, Yang, Kaixiang, Zhang, Guoming, Qian, Jiren, Zhu, Chengyuan
• Format: Conference Proceeding
• Language: English
• Published: IEEE 19.11.2022
• Subjects: Analytical models; Autocorrelation; Data models; egarch; heteroscedasticity; Load forecasting; natural gas load forecasting; Pipelines; Predictive models; sarima; Time series analysis
• DOI: 10.1109/YAC57282.2022.10023872

As low-carbon clean energy, natural gas plays an essential role in the energy transition, and the demand for natural gas is increasing rapidly. Natural gas load forecasting can not only understand the gas consumption characteristics of users, but also guide the formulation of natural gas pipeline network construction and dispatching plans. It is of great significance to improve the economic benefits of energy enterprises. The construction time of the natural gas pipeline network in Zhejiang Province is short, the amount of data for new users is small, and some external factors are difficult to obtain, which significantly affects the accuracy of load forecasting. To address the above issues, we conduct research on natural gas load forecasting, and design a time series forecasting model based on serial autocorrelation. Moreover, we have investigated the "leverage effect" in the financial field, and discussed the heteroscedasticity phenomenon. The principle and applicability of the SARIMA model and the EGARCH model are analyzed. The design ideas and modeling process of the fusion model are introduced. Finally, extensive experimental results show that the designed prediction model can achieve better performance.