A CSA-LSSVM Model to Estimate Diluted Heavy Oil Viscosity in the Presence of Kerosene

Viscosity is one of the properties that has important role in enhanced oil recovery processes, simulating reservoirs, and designing production facilities. Therefore, measurement and calculation of its accurate value is worthwhile. While the experimental methods for measurement of viscosity are expensive and time consuming, some credible correlations were developed to predict the viscosity with enough accuracy. For this purpose, in this study a balky data bank was gathered from open literature sources, and then one machine learning based approach called least square support vector machine (LSSVM) was utilized for prediction of heavy and extra-heavy crude oil viscosity. The parameters of proposed model were optimized by couple simulated annealing (CSA) optimization approach. The inputs of this model are temperature and kerosene mass fraction and the only output is viscosity.