• https://www.pvtsimnova.com/uploads/images/Top_images/modules.png

    PVT Simulation and Regression

PVT Simulation

These options allow for doing simulations representing experiments carried out in PVT laboratories. These experiments in turn represent a reservoir process or a separation process yielding information about the behavior of the fluid. The following PVT experiments may be simulated

  • Constant mass expansion
  • Constant volume depletion
  • Differential depletion
  • Saturation point
  • Separator test
  • Swelling test
  • Viscosity experiment
  • Multiple contact experiment
  • Slimtube
  • MMP

Up to five data sets for each type of experiment may be stored with each fluid composition. The user may define the stock tank conditions compliant with the actual PVT experiments. In the PVT simulation input menus, the entered temperatures and pressures are shown as default values when the experiments are simulated, and plots comparing experimental and simulated data are shown. However no experimental data are needed in order to perform the simulations. The input menus for storing the PVT data are arranged to comply with the usual standard of PVT reports, allowing for direct cut and paste of data from PVT reports available as soft copy. The output is arranged in the same way. Detailed output of physical properties and compositions at each pressure stage may optionally be shown. The extensive output allows the PVT laboratories to cross check the experimental data with simulated values. Using the Save Phase option, the phase composition at each stage may be stored in the database and used for subsequent calculations. This for example allows for studying the change in GOR of the liberated gas from a Constant Volume Depletion simulation as function of depletion pressure.

Experiments used to study miscibility of fluids are often time consuming and therefore expensive. Careful planning of which experimental conditions to use is therefore important. A multi contact experiment may be performed, or a slim tube experiment may be simulated to help narrow in on the pressures needed to get the part of the recovery curve of interest.

Compositional variation with depth in a reservoir may be important. Based on a sample composition and sample depth and PT-conditions, composition, pressure, GOR and a number of physical properties can be simulated within a specified depth interval along with location of the gas-oil contact if such exists. Simulations may be carried out isothermally or with application of a vertical temperature gradient.

If samples are available from multiple locations in a communicating fluid column, a regression can be made in order to match the observed compositional variation. The tuning is carried out with each components ideal gas enthalpy at the reference state as tuning parameters.

Regression

If PVT data measured in the laboratory are available for a fluid sample, simulation of those experiments, using the PVT simulation options, gives an overview of how well the models represent the measured data given the composition of the fluid sample. Regression is useful when the model parameters arising from the basic characterization procedure do not result in the desired agreement with PVT data. Data for up to five different experiments of the following types can be included in a regression

  • Constant mass expansion
  • Constant volume depletion
  • Differential depletion
  • Saturation point (including critical point)
  • Separator test
  • Swelling test
  • MMP
  • Viscosity experiment

The user may specify weights, expressing importance, to be assigned to each type of experimental data.

The regression minimizes the deviation between the measured data and the simulated results of the PVT experiments by varying a selection of model parameters. The progress can be followed as the object function, which is a measure for the deviation, is displayed during the regression. After completion, output and plots are presented showing the match of the measured data before and after regression. Values of the model parameters varied during regression are shown in the output in the form of the values before and after regression.

The regression may have an uncharacterized fluid (plus fraction or no-plus fraction fluid) or a characterized fluid as starting point. With an uncharacterized fluid as starting point, the characterization procedure is an integral part of the regression. The model parameters varied during regression are determined by PVTsim through an analysis of the included data and are typically Tc, Pc and Peneloux volume shift for the C7+ end of the fluid. The user has the option to view and modify the selection of parameters. The parameters are not varied directly but via manipulation of the characterization procedure, i.e. via the correlations yielding Tc, Pc and acentric factor (ω). The end result is a fluid characterized with a tuned characterization procedure. The regression can be performed in combination with the Normal or the Heavy oil C7+ characterization procedure.

With a characterized fluid as starting point, the model parameters are varied directly during the regression, and the user has full freedom in selection of model parameters. Any of the model parameters Tc, Pc, acentric factor (ω), Vc, Peneloux volume shift, Ωa , Ωb and temperature independent or temperature dependent binary interaction coefficients (kij) may be selected for any of the components in the fluid. Limitations on the change of the selected parameters can be modified. Parameters in the selected viscosity model (Corresponding States Principle or Lohrenz-Bray-Clark) may also be selected for tuning. The sensitivity of the object function with respect to changes in the model parameters is shown in the output. This can be valuable information when deciding which model parameters to select for tuning. With a characterized fluid as starting point, care should be taken not to violate too much the trends in the model parameter values established by the characterization procedure. It is therefore suggested to start out with a regression with the uncharacterized fluid as starting point to level out part of the differences between the measured data and the simulated results. If the obtained match is not satisfactory, fine-tuning can be done in a subsequent regression with the resulting characterized fluid as starting point.

If the tuned fluid is to be used in a third party compositional reservoir or flow simulator, it is often required to reduce the number of components through lumping to keep the computation time at an acceptable level. Reducing the number of components from the default number of approximately 20 to six or eight often results in less accurate simulation results. This introduces the need for combining the lumping of components with regression. The regression procedure automatically observes the applied lumping scheme. A special feature when lumping is the shifting of the critical point. If the critical temperature moves from one side of the reservoir or the wellhead temperature to the other, the fluid will in simulations change from a gas condensate to an oil or vice versa at the temperature in question. This causes severe problems. It is possible to supply the critical point as a special saturation point. If a high weight is assigned to this data point, the critical point can be prevented from shifting during regression.

Multiple samples may be available from different aerial locations and/or different depths in a reservoir, or fluids from different sources may meet in a merging network of pipelines. To carry out a compositional simulation utilizing all this information, all of these fluids should be represented by the same components and model parameters. The regression module may be used to characterize up to 50 different fluid compositions to the same pseudo-components with simultaneous regression. It is also possible to perform the regression using fluids that have previously been characterized (and tuned) to the same pseudo-components maintaining same pseudo-components and model parameters. All of the facilities mentioned for regression on a single fluid can be applied to this procedure, including a common lumping scheme and assignment of weights to the measured PVT data of each individual fluid. Fluids not having measured PVT data may be included to ensure that they get represented by the same components and model parameters as the tuned fluids