A data-driven model for the air-cooling condenser of thermal power plants based on data reconciliation and support vector regression. (25th January 2018)

• Record Type: Journal Article
• Authors: Li, Xiaoen; Wang, Ningling; Wang, Ligang; Kantor, Ivan; Robineau, Jean-Loup; Yang, Yongping; Maréchal, François
• Is Part Of: Applied thermal engineering. Volume 129(2018)
• Page Start: 1496
• Page End: 1507
• Publication Date: 2018-01-25
• Subjects: Air-cooling condenser -- Data-driven model -- Data reconciliation -- Support vector regression -- Thermal power plants; Heat engineering -- Periodicals; Heating -- Equipment and supplies -- Periodicals; Periodicals; 621.40205
• DOI: 10.1016/j.applthermaleng.2017.10.103
• Languages: English
• ISSNs: 1359-4311

Abstract

Highlights: A data-driven model of the air-cooling condenser by support vector regression. Data reconciliation for quality improvement of identified steady-state operating data. The derived model performs well under various ambient and operating conditions. Abstract: The performance of a direct air-cooling condenser under operation is rather complicated, as it is interactively affected by the operating conditions (e.g., the mode of air fans) and ambient conditions (e.g., temperature and wind speed). To understand the condenser's real performance under different situations, it is of great importance to investigate the relationship between the back pressure of the steam turbine and the condenser-related variables. However, direct analytical formulation or numerical simulation techniques both suffer from either inaccuracy or prohibitive computation time. In this paper, support vector regression method is applied to establish a data-driven model to express such a non-explicit relationship from the operating data. During raw-data processing, steady-state operation points are firstly identified by time-window method and properly sized for reasonable computational time. Then the reconciliation method is employed to improve the reliability and accuracy of measured data. The results show that the obtained data-driven model agrees well with the testing operation data under various boundary conditions, with a root mean square error of 0.81 kPa, a mean absolute error of 0.68 kPa and a correlation coefficient of 0.9675. It is also concluded that data reconciliation can increase the accuracy and stability of the data-driven model obtained with a reasonable computation time.