ConocoPhilips

This paper describes the statistical results from a large industry survey performed anonymously on mooring line failure on the behalf of the DeepStar. DeepStart project aimed at developing Mooring Integrity Management guidelines. The survey received strong support from the industry. The survey was complemented by an extensive review of accidents reported in the literature.
More than 120 observed failures or preemptive replacements were identified around the world enabling to perform some interesting statistical analysis to provide insight on such issues as:
• How are distributed the mooring failures around the world (Gulf of Mexico, Brazil, East Canada, North Sea, West Africa, Asia)?
• How did the number of failures vary over the last 15 years?
• What components (chain, wire, connector) are more prone to failure or that will require preemptive replacement.
• What is the historical expected life for components?
• What are the most common failure modes (corrosion, deployment, design analysis, fatigue, input data, manufacturing, motion, strength, wear/erosion, contact)?
• What are the associated frequencies?
• What is the distribution for the component lifetime?
Historical accident frequencies are inconsistent with service life expectations derived in accordance with conventional mooring design codes. In many regards these may be attributed to operating modes or localized mooring system detailing which are not explicitly considered in conventional design practice. These findings can improve the integrity of mooring systems.

The present paper proposes a novel non-intrusive model reduction approach based on Proper Orthogonal Decomposition (POD), the Discrete Empirical Interpolation Method (DEIM) and Radial Basis Function (RBF) networks to efficiently predict production of oil and gas reservoirs. Provided a representative set of training reservoir scenarios, either POD or DEIM allows for effectively projecting input parameters (e.g., permeability, porosity), states (e.g., pressure, saturations) and outputs (e.g., well production curves) into a much lower dimension that retains the main features contained in the simulation system. In this work, these projections are applied across multiple levels to be able to collapse a large number of spatiotemporal correlations. It is observed that these projections can be effectively performed at a large extent regardless of the underlying geological complexity and operational constraints associated with the reservoir model. The RBF network provides a powerful means for developing learning functions from input-output relationships described by the reservoir dynamics entailed by multiple combinations of inputs and controls. In order to achieve a high degree of predictability from the resulting reduced model, the RBF network exploits locality by a means of Gaussian basis functions that are maximal at the sampled point and decrease monotonically with distance. Compared to multilayer perceptron networks (i.e., traditional artificial neural networks) RBF networks require less training and are less sensitive to the presence of noise in the data. In this regard, POD or DEIM acts as a data filter that additionally aids at designing a more compact RBF network representation suitable for targeting fast reservoir predictions.

Abstract

In a fully-integrated reservoir and surface-facilities simulator, the overlapping multiplicative Schwarz method was chosen to solve the coupled system, using the perforated grid blocks as the overlapping layer. In many cases, it was found that this overlapping approach did not improve the global linear-solver performance, and the matrix for the extended-surface network, which included the perforated grid blocks, was much larger, denser, and lost its original tree structure, which taxed the linear solver for the network. Furthermore, the pressure solver for the reservoir domain was totally decoupled from the network domain. Such a loose coupling in the pressure solution led to disappointing performance. In most cases, the accuracy of the pressure solution across the reservoir and surface network directly determines the performance of the global linear solve, so it is crucial to define an appropriate global pressure matrix to represent the flow exchange between these two domains.

Taking advantage of new formulations for a generalized network model of wells and facilities, in which node-based variables, pressure, and component compositions are chosen as the primary variables, instead of mixed-node and connection-based variables, algebraic methods are designed to reduce the full-system matrix, involving pressure and component masses for the reservoir domain and pressure and component compositions for the network domain, to a pressure-only matrix. This global pressure matrix works as the first-stage preconditioning matrix in a two-stage solution method. For the reservoir domain, a widely-used IMPES-like reduction method was implemented. This paper focuses on methods to construct the pressure matrix for the network domain and coupling matrices between those two domains.

The present work aims at predicting space-time pressure solutions via a novel non-intrusive reduced order simulation model. The construction of lowdimensional spaces entails the combination of the Discrete Empirical Interpolation (DEIM) method with a suitable regressor such as artificial neural network to accurately approximate the pressure solutions arising in an IMPES formulation. Two basic assumptions are key in the present work: (a) physics invariance and, (b) stencil locality. The first one allows for coping with the curse of dimensionality associated with the training of a lowerdimensional surrogate model. The second assumption enables to significantly reduce the input parameter space and therefore, infer the global solution from local mass conservation principles. These assumptions are inspired in the discretization of PDEs governing the flow in porous media and serve as a powerful vehicle to generate physics-based surrogate models at a low computational cost. Hence, without explicit or little knowledge of simulation equations and numerical schemes, a sequence of pressure solutions or initial guesses can be obtained from inexpensive solutions of reduced order models (ROMs). Numerical examples are provided to illustrate the potentials of the present approach.

Summary

Recent advances in multiscale methods have shown great promise in modeling multiphase flow in highly detailed heterogeneous domains. Existing multiscale methods, however, solve for the flow field (pressure and total velocity) only. Once the fine-scale flow field is reconstructed, the saturation equations are solved on the fine scale. With the efficiency in dealing with the flow equations greatly improved by multiscale formulations, solving the saturation equations on the fine scale becomes the relatively more expensive part. In this paper, we describe an adaptive multiscale finite-volume (MSFV) formulation for nonlinear transport (saturation) equations. A general algebraic multiscale formulation consistent with the operator-based framework proposed by Zhou and Tchelepi (SPE Journal, June 2008, pages 267-273) is presented. Thus, the flow and transport equations are solved in a unified multiscale framework. Two types of multiscale operators--namely, restriction and prolongation--are used to construct the multiscale saturation solution. The restriction operator is defined as the sum of the fine-scale transport equations in a coarse gridblock. Three adaptive prolongation operators are defined according to the local saturation history at a particular coarse block. The three operators have different computational complexities, and they are used adaptively in the course of a simulation run. When properly used, they yield excellent computational efficiency while preserving accuracy. This adaptive multiscale formulation has been tested using several challenging problems with strong heterogeneity, large buoyancy effects, and changes in the well operating conditions (e.g., switching injectors and producers during simulation). The results demonstrate that adaptive multiscale transport calculations are in excellent agreement with fine-scale reference solutions, but at a much lower computational cost.