An Iterative Aggregation/Disaggregation Approach for the Solution of a Mixed-Integer Nonlinear Oilfield Infrastructure Planning Model

Susara A. van den Heever, Ignacio E. GrossmannIndustrial 2000 · DOI

TL;DR

This paper introduces a novel iterative algorithm for optimizing offshore oilfield infrastructure planning. By directly incorporating complex nonlinear reservoir behavior and using a combination of logic-based methods, bilevel decomposition, and time period aggregation, the approach significantly speeds up the solution process while maintaining solution quality.

What Problem Does This Paper Address?

Planning the development of offshore oilfields is a complex optimization problem. It involves making decisions about installing production platforms, well platforms, and wells, as well as determining their capacities and production schedules over decades. These decisions are complicated by:

  • Nonlinear Reservoir Behavior: The way oil and gas are produced from reservoirs is inherently nonlinear and depends on factors like pressure and cumulative production.
  • Discrete and Continuous Decisions: Many decisions are discrete (e.g., build a platform or not), while others are continuous (e.g., platform capacity).
  • Multiperiod Nature: Decisions made today impact future production and profitability over a long planning horizon.
  • Computational Intractability: Traditional optimization models for such problems become computationally prohibitive due to the sheer number of variables and constraints, especially when nonlinearities are handled directly.

Previous approaches often simplified reservoir behavior by using linear approximations, which can lead to inaccuracies. This paper aims to address these challenges by developing a method that handles nonlinearities directly and efficiently solves large-scale, multiperiod oilfield planning problems.

Methods

The core of this paper is an iterative aggregation/disaggregation algorithm designed to tackle the complexity of multiperiod MINLP (Mixed-Integer Nonlinear Programming) models for oilfield planning. Key components include:

  • Logic-Based Methods: The problem is formulated using disjunctive programming, which naturally captures the "if-then" logic of decisions (e.g., if a platform is installed, then certain operational constraints apply).
  • Bilevel Decomposition: The problem is broken down into an upper-level "Design Problem" (DP) that makes high-level infrastructure decisions and a lower-level "Operational Planning Problem" (OP) that determines the detailed schedule for a fixed design. This decomposition reduces problem size.
  • Convex Envelopes: To handle non-convexities introduced by nonlinear reservoir behavior, the authors replace linearizations of non-convex functions with convex envelopes. This ensures that the optimization algorithm can find a global optimum, rather than getting stuck in local optima.
  • Aggregation of Time Periods: To further reduce computational burden, especially for the upper-level design problem, time periods are aggregated into larger "super-periods." This simplifies the problem by reducing the number of time-dependent constraints.
  • Dynamic Programming Subproblem (PAI): A novel dynamic programming subproblem is introduced to intelligently determine the optimal way to aggregate time periods. This subproblem aims to create an aggregate problem that closely resembles the disaggregate problem, thereby improving the quality of the upper bound obtained from the aggregated design problem.

The overall algorithm iteratively refines the aggregation scheme and the solution by alternating between solving the aggregated design problem and the disaggregated operational planning problem.

Datasets

The paper uses data from a previously published oilfield planning problem (Iyer et al., 1998) for its examples. No new datasets are introduced.

Key Insights and Ideas

  • Directly Modeling Nonlinearities is Crucial: Approximating nonlinear reservoir behavior with linear constraints can lead to significant inaccuracies and suboptimal solutions. The paper demonstrates the benefit of incorporating these nonlinearities directly.
  • Convex Envelopes are Key for Global Optimality: When dealing with non-convexities, standard linearization techniques can cut off feasible regions and lead to suboptimal solutions. Convex envelopes provide a robust way to maintain global optimality.
  • Intelligent Aggregation Matters: Simply aggregating time periods can lead to poor upper bounds. The proposed dynamic programming subproblem (PAI) intelligently determines the aggregation scheme to minimize the gap between upper and lower bounds, leading to better estimates of the optimal solution.
  • Iterative Refinement Pays Off: The iterative nature of the algorithm, where the solution from one stage informs the next, allows for continuous improvement of both the solution and the aggregation strategy.

Key Results

  • Significant Speedup: The proposed iterative aggregation/disaggregation algorithm achieved an order of magnitude reduction in solution time compared to a full-space Outer Approximation (OA) algorithm. For a large industrial example, the proposed method solved the problem in 1423.4 CPU seconds, while the OA algorithm took 19385.9 CPU seconds.
  • Maintained Solution Quality: Despite the speedup, the proposed method maintained solution quality comparable to the full-space OA algorithm. The lower bounds obtained were similar or slightly better.
  • Improved Upper Bounds: The inclusion of the dynamic programming subproblem (PAI) for aggregation significantly reduced the gap between upper and lower bounds. For Example 1, the gap was reduced from 22.6% to 6.1%.
  • Effective Handling of Non-convexities: The use of convex envelopes successfully addressed non-convexities, leading to optimal solutions where linearization-only approaches yielded suboptimal ones.

Limitations and Future Work

  • More Complex Nonlinearities: The authors suggest incorporating more complex nonlinearities in future work.
  • Uncertainty: Dealing with uncertainty in reservoir simulations and future economic factors (like crude oil prices) is identified as a promising area for future research.
  • Sensitivity Analysis: Investigating the sensitivity of the solution to future crude oil prices is also proposed.

Summary updated

Abstract

A multiperiod mixed-integer nonlinear programming model for offshore oilfield infrastructure planning is presented where nonlinear reservoir behavior is incorporated directly into the formulation. Discrete decisions include the selection of production platforms, well platforms and wells to be installed/drilled, and the drilling schedule for the wells over the planning horizon. Continuous decisions include the capacities of the platforms and the production profile for each well in each time period. For the solution of this model, an iterative aggregation/disaggregation algorithm is proposed in which logic-based methods, a bilevel decomposition technique, the use of convex envelopes, and aggregation of time periods are integrated. Furthermore, a novel dynamic programming subproblem is proposed to improve the aggregation scheme at each iteration in order to obtain an aggregate problem that resembles the disaggregate problem more closely. A number of examples are presented to illustrate the performance of the proposed method.
BibTeX
@inproceedings{heever2000iterative,
  author = {Susara A. van den Heever and Ignacio E. Grossmann},
  title = {{An Iterative Aggregation/Disaggregation Approach for the Solution of a Mixed-Integer Nonlinear Oilfield Infrastructure Planning Model}},
  booktitle = {Industrial},
  year = {2000},
  doi = {10.1021/ie9906619},
}
Figure