Novel long short-term memory neural network considering virtual data generation for production prediction and energy structure optimization of ethylene production processes. (5th March 2023)

• Record Type: Journal Article
• Authors: Han, Yongming; Du, Zilan; Geng, Zhiqiang; Fan, Jinzhen; Wang, Youqing
• Is Part Of: Chemical engineering science. Volume 267(2023)
• Publication Date: 2023-03-05
• Subjects: Production prediction -- Energy structure optimization -- Virtual data generation -- Monte Carlo -- Long short-term memory neural network -- Ethylene production industry; Chemical engineering -- Periodicals; Génie chimique -- Périodiques; Chemical engineering; Periodicals; Electronic journals; 660
• DOI: 10.1016/j.ces.2022.118372
• Languages: English
• ISSNs: 0009-2509

Abstract

Graphical abstract: Highlights: Novel LSTM considering virtual data generation is proposed. Production prediction and energy structure optimization model is established. The prediction accuracy of the proposed model in ethylene industries is 96.57%. Carbon emission reduction can be achieved and energy-saving potential is 13.22%. Abstract: Production optimization and energy efficiency improvement can help realize the energy conservation and carbon emission reduction in the process industry. However, the amount of statistical data acquired in the process industry is relatively small, which is not conducive to production optimization modeling and analysis. Therefore, this paper proposes a novel production prediction and energy structure optimization model based on the long short-term memory neural network (LSTM) combining the Monte Carlo (MC) (MC-LSTM). The MC method can expand the real production data as the input of the LSTM model to realize the production prediction. At the same time, indicators of inefficient samples can be optimized for higher energy efficiency by analyzing the result of the MC-LSTM model. Finally, the proposed model is applied to predict the production and optimize the energy structure of ethylene plants in the process industry. The experiment shows that the prediction accuracy of the ethylene production process based on the proposed model is about 96.57%, which is better than other prediction models, and the energy-saving potential is 13.22% approximately.