Integration of LSSVM technique with PSO to determine asphaltene deposition

• A scaling-based model is presented for prediction of asphaltene precipitation.
• The model inputs include dilution ratio, molecular weight, and temperature.
• Least Square Support Vector Machine (LSSVM) combined with Particle Swarm Optimization (PSO) is utilized.
• Optimal system parameters are obtained by PSO.
• An acceptable match is noticed between measured and predicted values.

Asphaltene deposition is a recognized phenomenon in petroleum industry with undesirable outcomes so that it may lead to wellbore plugging and formation damage, resulting in a large amount of remedial costs to decrease its negative impacts on oil production. Therefore, it has attracted lots of research interests in the literature. In this study, an attempt is made to introduce the least square support vector machine (LSSVM) for prediction of asphaltene deposition. This technique with high capabilities which captures the complex nature of asphaltene could be inferred as a scaling model. As there is no a standard procedure to determine the main parameters of the LSSVM model, the particle swarm optimization (PSO) technique is employed to synchronously optimize the LSSVM parameters. The modeling results clearly demonstrate that the optimized LSSVM is able to handle the nonlinearities well and attain satisfactory results. The comparison of available predictive equations for asphaltene deposition confirms that the LSSVM technique linked with PSO exhibits higher robustness and greater precision with an R2 of 0.989 for the testing phase.