TY - JOUR
T1 - A Methodology for Designing Octane Number of Fuels Using Genetic Algorithms and Artificial Neural Networks
AU - Alboqami, Faisal
AU - van Oudenhoven, Vincent C. O.
AU - Ahmed, Usama
AU - Zahid, Umer
AU - Emwas, Abdul-Hamid M.
AU - Sarathy, Mani
AU - Abdul Jameel, Abdul Gani
N1 - KAUST Repository Item: Exported on 2022-04-26
Acknowledgements: Support received from the Interdisciplinary Research Center for Refining & Advanced Chemicals (CRAC), King Fahd University of Petroleum and Minerals (KFUPM), Saudi Arabia, under funded project INRC2104
PY - 2022/3/11
Y1 - 2022/3/11
N2 - Refineries often blend fuels in order to achieve mandated fuel properties based on international and domestic standards. One of the mandated properties for gasoline is Octane Number (ON). There are two forms of measuring ON: Research Octane Number (RON) and Motor Octane Number (MON). Although there have been efforts to optimize the blending operation in refineries, the implementations of advanced machine learning models are limited. The intention behind exploring and enhancing this field is to avoid undesirable scenarios, such as off-specification products and quality giveaways, as these scenarios overburden refineries' operational expenditures. The current work presents an innovative approach to predict the optimum fuel blending mechanism. This is accomplished through utilizing an integrated system composed of genetic algorithms (GA) and artificial neural networks (ANN). In addition, this work presents the polygonal algorithms that govern the optimization process. The system analyzes fuel inputs consisting of pure hydrocarbons, hydrocarbon-ethanol blends, and FACE (fuels for advanced combustion engines) gasoline-ethanol blends. There are ten pure hydrocarbons and six FACE gasoline's that are the base of 243 fuels used. These components are presented as multiple input streams to the blending stage, at which the system computes multiple recipes utilizing the GA. The system derives these recipes by capitalizing on the polygonal method. This method offers a systematic approach to design optimal fuel with the lowest addition of relatively high octane components. The system is evaluated using 243 blends. The produced recipes are designed to meet the desired antiknocking index (AKI). The coefficient of determination (R2) result is 0.99 for RON, MON, and AKI, respectively. Furthermore, the mean absolute error (MAE) values are 1.99, 1.23, and 1.40 for RON, MON, and AKI, respectively. These results signify the success of the integrated system and its significant potential impact in mitigating undesirable blending scenarios.
AB - Refineries often blend fuels in order to achieve mandated fuel properties based on international and domestic standards. One of the mandated properties for gasoline is Octane Number (ON). There are two forms of measuring ON: Research Octane Number (RON) and Motor Octane Number (MON). Although there have been efforts to optimize the blending operation in refineries, the implementations of advanced machine learning models are limited. The intention behind exploring and enhancing this field is to avoid undesirable scenarios, such as off-specification products and quality giveaways, as these scenarios overburden refineries' operational expenditures. The current work presents an innovative approach to predict the optimum fuel blending mechanism. This is accomplished through utilizing an integrated system composed of genetic algorithms (GA) and artificial neural networks (ANN). In addition, this work presents the polygonal algorithms that govern the optimization process. The system analyzes fuel inputs consisting of pure hydrocarbons, hydrocarbon-ethanol blends, and FACE (fuels for advanced combustion engines) gasoline-ethanol blends. There are ten pure hydrocarbons and six FACE gasoline's that are the base of 243 fuels used. These components are presented as multiple input streams to the blending stage, at which the system computes multiple recipes utilizing the GA. The system derives these recipes by capitalizing on the polygonal method. This method offers a systematic approach to design optimal fuel with the lowest addition of relatively high octane components. The system is evaluated using 243 blends. The produced recipes are designed to meet the desired antiknocking index (AKI). The coefficient of determination (R2) result is 0.99 for RON, MON, and AKI, respectively. Furthermore, the mean absolute error (MAE) values are 1.99, 1.23, and 1.40 for RON, MON, and AKI, respectively. These results signify the success of the integrated system and its significant potential impact in mitigating undesirable blending scenarios.
UR - http://hdl.handle.net/10754/676456
UR - https://pubs.acs.org/doi/10.1021/acs.energyfuels.1c04052
UR - http://www.scopus.com/inward/record.url?scp=85126571765&partnerID=8YFLogxK
U2 - 10.1021/acs.energyfuels.1c04052
DO - 10.1021/acs.energyfuels.1c04052
M3 - Article
SN - 0887-0624
JO - Energy & Fuels
JF - Energy & Fuels
ER -