Application of SVM to the Prediction of Water Content in Crude Oil

Water content of crude oil has always been an important indicator of evaluating the exploiting capacity of an oil field. Accurate rate of water content will optimize the production and decrease energy consumption. Due to the complicated working condition, large-scale experiments are designed and carried out in the simulation device of multiphase flow. After researching into the non-linear mapping relation between the frequency response of water content and its influencing factors, a prediction model of water content in crude oil about horizontal oil well based on SVM is proposed. The simulation results suggest that the SVM prediction model has higher prediction accuracy and stronger capability of generalization compared with the BP neural network. It will provide a promising theoretical and practical perspective for the explanation and prediction of the data acquired from the oil field.