Accuracy of Modeling Multiphase Flow in Oil/Gas Upstream Facilities
EnSys Yocum has often been asked what type of accuracy can be achieved in simulating multiphase flow through oil wells, gas-oil separators, and surface flowlines. This blog entry discusses the accuracy of the EnSys Yocum simulators WELLSIM, GOSPSIM, and PRODSIM and the various forms of error encountered in modeling upstream facilities.
When working in the Middle East, in conditions with engineered production flow tests and close monitoring by engineers, we consistently achieved 3-10 percent accuracy over a range of well/flowlines and crude types. Wells were modeled with higher accuracy (3-5 percent) whereas the flowlines tended to have higher levels of uncertainty (5-10 percent range). The flow regime has an effect on the level of uncertainty with homogenous flow tending to represent the higher accuracy simulations, stratified flow for the middle accuracies, and slug flow representing the most uncertainty. To simulate these systems in the Middle East and obtain these accuracies, we were using Core Laboratory Assays of Equilibrium Flash and Differential Expansion which provided information at standard/alternate conditions and above/below the bubble point). We also possessed information on gas-oil ratio, mixture density, liquid density, gas density, and liquid viscosity. The facilities simulated comprised dozens of wells and miles of surface flowlines over mountainous terrain. The underlying laboratory studies of the reservoir fluid properties were accurate to 3-6 percent. We also had highly reliable downhole information (static reservoir pressure/temperature) via Schlumberger wireline tool. Two thirds of the oil well pressure drop is across the reservoir pay and therefore accurate downhole tests were key to modeling with high accuracy. Additionally, we received accurate surface pressure/temperature information from gauges.
The “short-cut PVT approach” built into the EnSys Yocum simulators enable a user to select an assay already in the EnSys Yocum database and then adjust for (primarily) GOR. This implies a higher level of uncertainty albeit with a quicker/more convenient approach covering a wide range of crude types where the client has not supplied a field-specific assay. Beyond the PVT, we see uncertainties in the aforementioned pressure/temperature readings (both downhole and wellhead/surface) and also representation of (configuration data for) the underlying system being modeled. Since many operating companies today possess databases describing well configurations and, if this system is ultimately configured to draw from the database, the error in respect to configuration data should be minimal. After simulating various wells, surface flowlines, and gas-oil separators using our “short-cut approach” with regard to PVT selection and minimum configuration data input, we expect uncertainties between 5-20%. (This has been the case with several simulations undertaken recently in the Middle East).
Two other types of uncertainties are rapid flow pulsations and transmission signals. We have encountered ‘pulsation’ flowing phenomena in a couple of instances before and this obviously adds significantly to the uncertainty (the model is based on steady state flow multiphase flow). In assessing transmission signal, it is difficult to determine the upper limit of uncertainty. Ultimately, garbage in will produce garbage out just like with any simulator.