Real-time monitoring of the oil shale pyrolysis process using a bionic electronic nose. (1st April 2022)

• Record Type: Journal Article
• Authors: Zhao, Rongsheng; Kong, Cheng; Ren, Luquan; Sun, Youhong; Chang, Zhiyong
• Is Part Of: Fuel. Volume 313(2022)
• Publication Date: 2022-04-01
• Subjects: Oil shale -- Pyrolysis process -- Bionic electronic nose -- Real-time monitoring -- Algorithm model; Fuel -- Periodicals; Coal -- Periodicals; Coal; Fuel; Periodicals; 662.6
• DOI: 10.1016/j.fuel.2021.122672
• Languages: English
• ISSNs: 0016-2361

Abstract

Highlights: The regular evolution of the pyrolysis product with changing heating rate allowed the oil shale pyrolysis process to be monitored in real time by a bionic electronic nose. We qualitatively checked whether the oil shale had entered the hydrocarbon generation stage with a recognition rate of 91.13%. Quantitative recognition of the Easy%Ro evolution of hydrocarbon generation was performed with an R 2 reaching 0.95. Abstract: Accurate and rapid prediction of thermal maturity during oil shale exploitation is important to optimize pyrolysis processes and decrease production costs. Then, a bionic electronic nose (BEN) was used for the first time for the real-time monitoring of the oil shale pyrolysis process. The results show that the calculated Easy%Ro change, unlike that of the measured vitrinite reflectance (%Ro ), was small (±0.004) in decomposition stages (dewatering, hydrocarbon generation stage and inorganic matter decomposition) at the different heating rate. This time–temperature-related parameter made it a perfect intermediary to associate the thermal maturation process with the BEN signal during oil shale pyrolysis. The monitoring of the pyrolysis process of oil shale in real time was divided into two steps: 1) qualitatively checking whether the oil shale had entered the hydrocarbon generation stage, and 2) when entering this stage, quantitatively predicting the Easy%Ro evolution. Combined with different feature extraction methods, a support vector machine (SVM) as a classifier was used to complete the first step, which had the best recognition rate of 91.13%. For the second step, random forest (RF) was applied to quantitatively measure the Easy%Ro values, with an R 2 reaching 0.95. During the verification stage, the established integration model was used to recognize a heating rate, which, between the trained heating rate in the algorithm model, performed much better in the first step (SVM, 92.96%) than in the second step (RF, 0.57), and those results were given in 38 to 125 s. Thus, the BEN technique can be applied in the real-time detection of oil shale exploration.