Zhu, Jianjun (University of Tulsa) | Zhu, Haiwen (University of Tulsa) | Cao, Guangqiang (PetroChina Company Ltd.) | Banjar, Hattan (Saudi Aramco) | Peng, Jianlin (University of Tulsa) | Zhao, Qingqi (University of Tulsa) | Zhang, Hong-Quan (University of Tulsa)
As the second most widely used artificial lift method in petroleum industry, ESPs help maintain or increase flow rates by converting kinetic energy to hydraulic pressure. During the entire life of an oilfield, water is invariably produced with crude oil. As the field ages, the water cut in production increases. Due to high shear force inside rotating ESPs, the oil-water emulsions may form, which can be stabilized by natural surfactants or fine solids existing in the crude oil. The formation of emulsions during oil production create high viscous mixture, resulting in costly problems and flow assurance issues, such as pressure drop increase and production rate lost. This paper, for the first time, proposes a new mechanistic model for predicting oil-water emulsion rheology and its effect on the boosting pressure in ESPs. The model is validated with experimental measurements with an acceptable accuracy.
The new mechanistic model starts from Euler equations for centrifugal pump, and introduces a conceptual best-match flowrate
The mechanistic model-predicted ESP water performance curves are found to match the catalog curves perfectly. With high-viscosity fluid flow, the model predictions of ESP boosting pressure agree well with the experimental data. For most calculation results within medium to high flow rates, the prediction error is less than 15%. With oil-water two-phase flow, the proposed rheology model predicts the effective viscosities of emulsions match testing results with 10% prediction error. The inversion points, at which the continuous phase changes from oil to water as water cut increases, are also predicted. The predictions of ESP boosting pressure under oil-water emulsion flow by coupling the mechanistic model and emulsion rheology model are comparable with experimental results.
Zhu, Jianjun (University of Tulsa) | Cao, Guangqiang (PetroChina Company Ltd.) | Tian, Wei (PetroChina Company Ltd.) | Zhao, Qingqi (University of Tulsa) | Zhu, Haiwen (University of Tulsa) | Song, Jie (PetroChina Company Ltd.) | Peng, Jianlin (University of Tulsa) | Lin, Zimo (University of Tulsa) | Zhang, Hong-Quan (University of Tulsa)
Plunger lift has been widely used in unconventional gas wells to remove liquid accumulation from the well.. Production surveillance provides large amount of data of production process and normal and abnormal operations, which can be used in machine learning (ML) and Artificial Intelligence (AI) to develop algorithms for anomaly diagnosis and operation optimization. However, in the surveillance data the majority is related to daily operation and the data of failure cases are rare. Also the failure cases may not be repeatable and many failure case signatures are not available until they happen. Large data size of anomaly cases are needed to improve the ML model accuracy. Dynamic simulation of the plunger lift process offers an alternative way to generate synthetic data on the specified anomalies to be used to train the ML model. It also helps better understand the trends reflected in the surveillance data and their root causes.
From the available surveillance data of gas wells equipped with plunger lift, the simultaneous measurements of different parameters at different points in a production system with normal and abnormal occurrences can be analyzed and the correspondent trends/signatures can be identified. The typical signatures that conform to pre-determined anomalous patterns can be obtained. Using a commercial transient multiphase flow simulator, the actual field data of tubing/casing pressures can be matched through a tuning process. Trial-and-error is needed to improve the dynamic plunger lift model so that a good agreement with the production data can be achieved by adjusting the reservoir performance, plunger parameters or surface pipeline boundary conditions. Following the validation under different flow conditions, synthetic datasets for various operational and flow conditions can be generated by performing parametric studies. Unlike the field data, the synthetic data from the dynamic simulations mainly comprise anomaly signatures (e.g. tubing rupture, missed arrival of plunger, etc.), which can be added to the ML data pool to reduce the data covariance and increase independency.
In the petroleum industry, well testing is a common practice that consists of wellbore pressure, temperature and flow rates data acquisition to estimate parameters that govern the flow in porous media. Injection-falloff testing is particularly important for offshore reservoirs, especially for the oil reserves that contain high carbon dioxide and sulfur content. In this environment, a conventional well test in an exploratory well should not be run in order to avoid discarding high concentrations of these gases to the atmosphere. Therefore, there is a need for developing techniques for analyzing pressure data from injection-falloff tests. In this work, we have developed an approximate semi-analytical solution for wellbore pressure response during gas injection and falloff well tests in reservoirs containing oil and gas with complex composition by applying the Thompson and Reynolds steady-state theory. For the injection period, we first determine the overall concentrations distributions from a system of hyperbolic conservation equations using the method of characteristics (MOC), assuming a one-dimensional homogeneous reservoir with incompressible fluids and constant molar density, and neglecting capillary, gravity effects, volume changing on mixing and diffusion. During the falloff stage, it is assumed that there is no phase nor concentration movement in the reservoir, which is reasonable as we neglect capillary pressure, diffusion, gravity force and fluid compressibilities. Once we have the concentration profiles in the reservoir, we can calculate the total mobility distributions and then integrate the pressure gradient given by Darcy's law to find the wellbore pressure response. The semi-analytical approximate solution obtained was validated against the commercial numerical simulator STARS from CMG. After validation, the developed model was used as a forward model to estimate absolute permeability and skin factor by history matching noisy data obtained from the numerical simulator mentioned.
While distributed temperature sensing (DTS) has become a commonly used tool in reservoir studies, the technology has not seen widespread use in SCAL projects. Most core-scale experiments attempt to control temperature at a constant value rather than monitor temperature changes within a sample during a test. High-resolution temperature arrays are available that measure changes in temperature of 0.1°C at 1-mm resolution. The optical backscatter reflectance (OBR) fiber senses both temperature and strain that can be separated through experiment design and signal processing. These OBR fibers are sensitive enough to monitor temperature changes associated with endo- and exothermic chemical reactions associated with mineral dissolution/precipitation, or fluid-front movements in steam-assisted gravity drainage of heavy-oil tests. An example of the use of a distributed temperature array is in the monitoring of natural-gas-hydrate formation and dissociation in a sandpack as CO2 is exchanged with the naturally occurring CH4 in the hydrate structure. A fiberoptic array was placed within a narrow-diameter PEEK tube as the sandpack was constructed. The PEEK tube held the fiber optic in place so that the sensed signal was temperature only and did not include any strain effects. The OBR was set up to acquire a temperature array every 30 seconds during the test at 5-mm spacings. The core holder was placed in a MRI instrument that provided additional spatial information on hydrate formation during the test that was compared with the OBR results. Initial hydrate formation resulted in a several degrees increase in temperature at the inlet end of the cell that with time, progressed down the length of the cell. The temperature array and MRI images both showed the nonuniform nature of hydrate formation and subsequent dissociation of the hydrate when N2 was injected into the cell as a permeability enhancement step. The faster response of the OBR array compared to the time required to acquire MRI images provided additional detail in the sequence of hydrate formation and dissociation during CH4-CO2 exchange. The limitation to the OBR array was that it only sensed temperature fluctuations proximal to the fiber as a function of the hydrate system’s thermal conductivity.
An XFEM-EDFM scheme and associated monolithic solution method are proposed to model time-dependent poromechanics and two-phase flow. Fractures are modeled as interfaces with displacement discontinuities. The contact forces are treated using Lagrange Multipliers. A number of numerical tests are performed to investigate the Newmark scheme's accuracy and cases for wave propagation in poroelastic and natural fracture media are implemented to evaluate computational efficiency. We apply the method to model seismic data from hydraulic fracture network. Empirical results validate the Newmark scheme accuracy as well as computational efficiency and localization of newton update in seismic field is necessary for the further application. The synthetic model of multiple hydraulic stages illustrates the effect of flow coupling and newly generated fractures on the microseismic field. The model is applied to simultaneously assimilate well performance and microseismic observations, thereby informing about the causal event dynamics.
Fracture propagation (FP) occurs in extensive applications including hydraulic fracturing, underground disposal of liquid waste, CO2 sequestration etc. It is crucial to develop a simulator that is able to reflect physics behind FP and capture the FP path. This work is an extension of the previous developed model (Ren et al. (2018), Ren & Younis (2018)) to the simulation of FP. One of the remarkable benefits using the coupled XFEM-EDFM scheme allows FP free of the remeshing. In this work, the onset of FP is controlled by a single parameter, the equivalent stress intensity factor (SIF). A domain integral method, J integral is applied to extract the SIF information. A time marching scheme is performed to ensure the SIF criterion satisfied everytime fracture propagates. The developed simulator is verified by the analytical solutions and shows the capability of FP simulation in poroelastic materials.
Solving a large-scale optimization problem with nonlinear state constraints has proven to be challenging when adjoint gradients are not available for computing the derivatives needed in the basic optimization algorithm employed. Here, we present a methodology for the solution of an optimization problem with nonlinear and linear constraints where the true gradients that cannot be computed analytically are approximated by ensemble-based stochastic gradients based on an improved stochastic simplex approximate gradient (StoSAG). For the most part, our discussion is focused on the application of our procedure to waterflooding optimization where the optimization variables are the well controls and the cost function is the life-cycle net present value (NPV) of production. The optimization algorithm used for solving the constrained optimization problem is sequential quadratic programming (SQP) with constraints enforced using the filter method. We introduce modifications to StoSAG that improve its fidelity, i.e., the improvements give a more accurate approximation to the true gradient (assumed here to equal the gradient computed with the adjoint method) than the approximation obtained using the original StoSAG algorithm. The improvements to the basic StoSAG vastly improve the performance of the optimization algorithm; in fact, we show that if the basic StoSAG is applied without the improvements, then SQP may yield a highly suboptimal result for optimization problems than many nonlinear state constraints involve.
Multistage hydraulic fracturing of a horizontal well in an unconventional reservoir tends to induce a complex fracture network (CFN) which is challenging to characterize by conventional methods. In this work, we develop a fracture characterization workflow to estimate the geometric configuration and fracture properties of a CFN by assimilating microseismic event data and production data, sequentially.
A novel stochastic fractal model, that is consistent with rock physics and outcrop observations, is developed in order to generate realizations of the complex fracture network. In the first stage of the two-stage assisted history matching workflow, we estimate the parameters of the stochastic fractal model (fracture intensity, average fracture length, orientation and fracture distribution) by using a genetic algorithm to history match data for the locations of microseismic events. In the second stage, the production data from the shale reservoir are assimilated by the ES-MDA algorithm to estimate the stimulated reservoir volume (SRV) and its average permeability, fracture permeability, aperture and porosity. In the unconventional shale gas reservoir simulator used as the forward model, large-scale fractures are modeled via the embedded discrete fracture model (EDFM) and a dual-porosity, dual-permeability (DP-DK) model is used for modeling the SRV and small scale fractures. The simulator includes Knudsen diffusion and the Langmuir adsorption/desorption model.
For validation, we consider a synthetic shale gas reservoir with a horizontal well that has been stimulated by multistage hydraulic fracturing. A particular realization of the variables that describe the reservoir model is used to generate observed data for microseismic events and production rates. The parameters to be adjusted to match the observed microseismic events are the expected values of the length, orientation and intensity of the distribution of the natural fractures and the fractal pattern. Results show that we obtain good estimates of the expected value of natural fracture length, orientation, intensity and fracture distribution by history matching observations of locations of microseismic events. These estimates provide an updated stochastic fractal model for the configuration of CFN. The history-matched fractal model is used to generate an ensemble of fracture distributions consistent with microseismic data as candidate fracture configurations when estimating fracture properties by matching production data. We obtain much better history matches, future performance predictions, estimates of stimulated reservoir volume and its average permeability and estimates of fracture permeability, porosity and aperture when we match both microseismic and production data than we only match production data. When both seismic and production data are matched for synthetic cases and parameters are properly scaled, the true values of parameters and reservoir performance predictions are within the P25-P75 confidence intervals calculated from the ensemble of history matched models in virtually all cases.
In practice, the proper characterization of the CFN and reservoir properties should be useful for placing new wells and designing fracture treatments.
Geophysical Reservoir Monitoring GRM systems such 4D seismic are increasingly used in the oil and gas industry because they provide unique and useful information on fluid movement within the reservoir. This information is relevant for many reservoir management decisions; including new well placement, well intervention, and reservoir model updating.
Unfortunately, it has been difficult to estimate the value creation of any data acquisition scheme due to the fact that a multidisciplinary approach is required to model the value that future measurements will imply in future decisions. This assessment requires a common decision making simulation frame work that can integrate the input from geo-modelers, geophysicist and reservoir engineers.
This work presents an example of how a Close Loop Reservoir Management (CLRM) simplification can be used as a framework for simulating NPV changes due to assimilation of production and saturations in a simple toy model. It combines state-of-the-art data assimilation and uncertainty modeling methods with a robust optimization genetic algorithm to calculate NPV improvements due to model update and its relationship with the NPV obtained from the synthetic reservoir.
In this context a simple synthetic model is presented. It recreates a segment of green field under a strong aquifer influence with two discovery wells. The reservoir development requires the selection of 4 well locations at fixed drilling times. The development strategy selection is obtained with the use of a genetic algorithm within the CLRM framework. Subsequently two cases are presented: one of assimilating only production after the first two wells have been drilled, just before deciding the locations of the last two wells; and a second case, in which production and saturation are assimilated at the same time. The saturation map assimilated is assumed to be output of a 4D seismic acquisition. The model update imposes the need of optimally relocate the last two wells which results in a NPV change.
The results show how the obtained NPVs is incremented by the relocation of the last two wells in both cases. A bigger increment is obtained when both, production and saturation are assimilated. In addition, the ensemble improved its forecast capability the most, when saturation assimilation is included. Nevertheless, the ensemble expected NPV decreases after assimilation from the value obtained from the first development strategy optimization; this indicates an optimistic early NPV valuation due to the initial ensemble distributions spread.
The study presents an asset simulation framework that could be used to evaluate data acquisition investments through the systematic modeling of reservoir uncertainties with in a decision oriented focus. This could include the inclusion of additional uncertain model parameters, the insertion of water injector and well conversions, the assimilation of saturations at different intervals, the change on the quality of the saturation maps assimilated, in addition to sensitivity studies of other economic constrains.
Accompanied with liquid condensation, natural gas production wells suffer from liquid loading if the gas flow rate is insufficient to carry liquids to the surface. With continuous production, the reservoir pressure decreases due to reservoir depletion, resulting in decrease of gas flow rate and inability to carry liquid upward. Then, the produced liquid accumulates in the well bottom and creates a static liquid column, adding a backpressure against reservoir pressure and reducing gas flow rates until the well production ceases. Due to many advantages, such as low operation cost and prevention of paraffin deposition along wellbore, plunger lift has been widely used in gas wells for the removal of liquid column and rescuing dying gas wells from liquid loading.
The existing plunger lift models in literature are imperfect due to either limited field applications or oversimplified assumptions, which lead to considerable prediction errors. Starting from