Investigation of Asphaltene Precipitation Using Response Surface Methodology Combined with Artificial Neural Network

• Main Authors: Zeinab Hosseini-dastgerdi, Saeid Jafarzadeh-Ghoushchi
• Format: Article
• Language: English
• Published: University of Tehran 2019-01-01
• Series: Journal of Chemical and Petroleum Engineering
• Subjects: artificial neural network asphaltene; desirability; precipitation; response surface methodology
• Online Access: https://jchpe.ut.ac.ir/article_73622.html

The precipitation of asphaltene, one of the components of oil, in reservoirs, transfer lines, and equipment causes many problems. Accordingly, researchers are prompted to determine the factors affecting asphaltene precipitation and methods of avoiding its formation. Predicting precipitation and examining the simultaneous effect of operational variables on asphaltene precipitation are difficult because of the multiplicity, complexity, and nonlinearity of factors affecting asphaltene precipitation and the high cost of experiments. This study combined the use of response surface methodology and the artificial neural network to predict asphaltene precipitation under the mutual effects of various parameters. The values of such parameters were determined to reach the minimum amount of precipitation. We initially selected the appropriate algorithm for predicting asphaltene precipitation from the two neural network algorithms. The outputs of designed experiments in response surface methodology were determined using the optimum algorithm of the neural network. The effects of variables on asphaltene precipitation were then investigated by response surface methodology. According to the results, the minimum precipitation of asphaltene achieved at zero mole percent of injected nitrogen and methane, 10–20 mole percent of injected carbon dioxide, asphaltene content of 0.46, the resin content of 16.8 weight percent, the pressure of 333 psi, and temperature of 180 . Results showed that despite the complexities of asphaltene precipitation, the combination of artificial neural network with response surface methodology can be successfully used to investigate the mutual effect of different variables affecting asphaltene precipitation.