4.7 Article

Multi-feature-scale fusion temporal convolution networks for metal temperature forecasting of ultra-supercritical coal-fired power plant reheater tubes

Journal

ENERGY
Volume 238, Issue -, Pages -

Publisher

PERGAMON-ELSEVIER SCIENCE LTD
DOI: 10.1016/j.energy.2021.121657

Keywords

Multi-feature-scale fusion temporal convolution networks; Metal temperature forecasting; Feature construction; Spatial partition model; Feature fusion

Funding

  1. Natural Science Foundation of Guangxi Province Grant [AD19245001, 2020GXNSFBA159025]

Ask authors/readers for more resources

A multi-feature-scale fusion temporal convolution network prediction model is proposed in this paper for accurate real-time prediction of the metal temperature of reheater tube in a thermal power plant. The model demonstrates strong feature extraction and nonlinear function fitting capabilities, with innovative methods to improve accuracy. The proposed model outperforms 20 other prediction models, achieving higher accuracy in predicting the metal temperature of the reheater tube.
The ultra-supercritical coal-fired power plant has higher reheater steam temperatures. To reduce the possibility of overheating operation and decrease the explosion of reheater, the operating temperature of reheater should be accurately monitored in real-time. Through the real-time prediction of the temperature distribution of the reheater tube, the alarm threshold is set to prevent tube burst and other faults. To accurately predict the metal temperature of the reheater tube in real-time, a multi-feature-scale fusion temporal convolution network prediction model is proposed in this paper. The model has a strong ability of feature extraction and nonlinear function fitting, and can infinitely approximate the mapping relationship between input and output data. Innovative methods feature construction, spatial partition and feature fusion, are proposed to improve the accuracy. Original data are collected from one thermal power plant reheater in the Guangdong province of China. The multi-feature-scale fusion temporal convolution network is applied to study the real-life data from March 2020 to May 2020. Compared with 20 prediction models such as random forest, the proposed model has higher accuracy. The mean absolute percentage error result of the proposed model is smaller 0.16% than the second smallest model and 25.70% smaller than the largest model. (c) 2021 Elsevier Ltd. All rights reserved.

Authors

I am an author on this paper
Click your name to claim this paper and add it to your profile.

Reviews

Primary Rating

4.7
Not enough ratings

Secondary Ratings

Novelty
-
Significance
-
Scientific rigor
-
Rate this paper

Recommended

No Data Available
No Data Available