Prediction of Gas/Water Relative Permeability Using the GWO-LSSVM Model Under HTHP Condition

The most common methods of predicting a gas/water two-phase relative permeability include experimental measurement, theoretical simulation, and numerical calculation. The experimental method is relatively accurate but time-consuming and laborious. The method of theoretical modeling involves a complicated solving process, and the range of application of the numerical calculation method is comparatively narrow. In this work, based on the least square support vector machine (LSSVM) combined with the gray optimization algorithm (GWO), the prediction model (GWO-LSSVM) of the gas/water phase permeability is established. The model was validated by the gas-water two-phase relative permeability data obtained by a non-steady-state method from a gas reservoir under high pressure/high temperature (133–152.5°C, 50.3–55 MPa) conditions. The irreducible water saturation (Swc ), absolute permeability (K), porosity (j), and gas saturation (Sg ) have been set as the input parameters. The water phase relative permeability (Krw) and the gas phase relative permeability (Krg) have been set as output variables. The training data of water phase and gas phase relative permeability were 147 and 141, respectively. The prediction data were 32 and 58, respectively. The results show that the absolute relative deviation (AARD) of Krw for the presented GWO-LSSVM model is 3.69%, while that of Krg is 2.92%, showing that the model can accurately predict the gas-water relative permeability at high temperature and high pressure. The method provides a new and effective instrument for estimating the gaswater two-phase relative permeability parameters in the development of gas reservoirs.