Tuning of parameters in a two-phase flow model using an ensemble Kalman filter

Lorentzen, Rolf Johan; Nævdal, Gunnar; Lage, Antonio

doi:10.1016/s0301-9322(03)00088-0

Cited by 67 publications

(42 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Modeling of UBD operations and MPD scenarios handling influx requires a multiphase model. A popular model in the literature is the Drift-Flux Model (DFM) [5], [6], [7]. The drift flux model is a set of first order nonlinear hyperbolic partial differential equations (PDE).…”

Section: Introductionmentioning

confidence: 99%

State and Parameter Estimation of a Drift-Flux Model for Underbalanced Drilling Operations

Nikoofard

Aarsnes

Johansen

et al. 2017

IEEE Trans. Contr. Syst. Technol.

View full text Add to dashboard Cite

Abstract-We consider a drift-flux model (DFM) describing multiphase (gas-liquid) flow during drilling. The DFM uses a specific slip law which allows for transition between single and two phase flows. With this model, we design Unscented and Extended Kalman Filters (UKF and EKF) for estimation of unmeasured state, production, and slip parameters using real time measurements of the bottom-hole pressure, outlet pressure, and outlet flow-rate. The OLGA high-fidelity simulator is used to create two scenarios from Under-Balanced Drilling (UBD) on which the estimators are tested: A pipe connection scenario and a scenario with a changing production index. A performance comparison reveals that both UKF and EKF are capable of identifying the production indices of gas and oil from the reservoir into the well with acceptable accuracy, while the UKF is more accurate than the EKF. Robustness of the UKF and EKF for the pipe connection scenario is studied in case of uncertainties and errors in the reservoir and well parameters of the model. It is found that these methods are very sensitive to errors in the reservoir pore pressure value. However, they are robust in the presence of error in the liquid density value of the model.

show abstract

Section: Introductionmentioning

confidence: 99%

State and Parameter Estimation of a Drift-Flux Model for Underbalanced Drilling Operations

Nikoofard

Aarsnes

Johansen

et al. 2017

IEEE Trans. Contr. Syst. Technol.

View full text Add to dashboard Cite

show abstract

“…In reservoir history-matching applications, DA has been used to update the dependent variables of multiphase flow models, such as pressure and saturations, and as an inverse modelling tool to "condition" model parameters, such as porosity and permeability, based on the observed data (e.g. Lorentzen et al 2003, Gu & Oliver 2005, Skjervheim et al 2011, Emerick & Reynolds 2013.…”

Section: Introductionmentioning

confidence: 99%

On the importance of the heterogeneity assumption in the characterization of reservoir geomechanical properties

et al. 2016

View full text Add to dashboard Cite

SUMMARYThe geomechanical analysis of a highly compartmentalized reservoir is performed to simulate the seafloor subsidence due to gas production. The available observations over the hydrocarbon reservoir consist of bathymetric surveys carried out before and at the end of a ten-year production life. The main goal is the calibration of the reservoir compressibility c M , i.e., the main geomechanical parameter controlling the surface response. Two conceptual models are considered: in one (a) c M varies only with the depth and the vertical effective stress (heterogeneity due to lithostratigrafic variability); in another (b) c M varies also in the horizontal plane, that is, it is spatially distributed within the reservoir stratigraphic units. The latter hypothesis accounts for a possible partitioning of the reservoir due to the presence of sealing faults and thrusts that suggests the idea of a block heterogeneous system with the number of reservoir blocks equal to the number of uncertain parameters. The method applied here relies on an ensemble-based data assimilation (DA) algorithm (i.e., the Ensemble Smoother, ES), which incorporates the information from the bathymetric measurements into the geomechanical model response to infer and reduce the 2 C. Zoccarato, D. Baù, F. Bottazzi, M. Ferronato, G. Gambolati, S. Mantica, & P. Teatini uncertainty of the parameter c M . The outcome from conceptual model (a) indicates that DA is effective in reducing the c M uncertainty. However, the maximum settlement still remains underestimated, while the areal extent of the subsidence bowl is overestimated.We demonstrate that the selection of the heterogeneous conceptual model (b) allows to reproduce much better the observations thus removing a clear bias of the model structure.DA allows significantly reducing the c M uncertainty in the five blocks (out of the seven) characterized by large volume and large pressure decline. Conversely, land subsidence can constrain only partially the partitions that marginally contributes to the cumulative displacements of the seafloor.

show abstract

“…compared and evaluated the performance of the extended Kalman filter, the ensemble Kalman filter and the unscented Kalman filter based on a low order model to estimate the states and the production index (PI) in UBD operation. Lorentzen et al (2003) designed an ensemble Kalman filter based on a drift-flux model to tune the uncertain parameters of a two-phase flow model in the UBD operation. Vefring et al (2003Vefring et al ( , 2006 compared and evaluated the performance of the ensemble Kalman filter and an off-line nonlinear least squares technique utilizing the Levenberg-Marquardt optimization algorithm to estimate reservoir pore pressure and reservoir permeability during UBD while performing an excitation of the bottom-hole pressure.…”

Section: Introductionmentioning

confidence: 99%

Evaluation of Lyapunov-based Adaptive Observer using Low-Order Lumped Model for Estimation of Production Index in Under-balanced Drilling

2015

View full text Add to dashboard Cite

A distributed drift-flux model and a low-order lumped model describing a multiphase (gas-liquid) flow in the well during Under-Balanced Drilling (UBD) has been presented. This paper presents a novel nonlinear adaptive observer to estimate the total mass of gas and liquid in the annulus and production constant of gas and liquid from the reservoir into the well during UBD operations. Furthermore, it describes a joint unscented Kalman filter to estimate parameters and states for both the distributed drift-flux and lumped model by using real-time measurements of the choke and the bottom-hole pressures. The performance of the adaptive observers are evaluated for typical drilling scenarios. The results show that all adaptive observers are capable of identifying the production index, although the adaptive observers based on the low-order lumped model achieves better convergence rate than adaptive observer based on the drift-flux model. The results show that the LOL model is sufficient for the purpose of estimating the production parameters.

show abstract

Tuning of parameters in a two-phase flow model using an ensemble Kalman filter

Cited by 67 publications

References 21 publications

State and Parameter Estimation of a Drift-Flux Model for Underbalanced Drilling Operations

State and Parameter Estimation of a Drift-Flux Model for Underbalanced Drilling Operations

On the importance of the heterogeneity assumption in the characterization of reservoir geomechanical properties

Evaluation of Lyapunov-based Adaptive Observer using Low-Order Lumped Model for Estimation of Production Index in Under-balanced Drilling

Contact Info

Product

Resources

About