Production optimization can play a major role in increasing recovery and decreasing operation cost. In many oilfields, the geology, production operations, and their related constraints are very complex. These complexities can complicate the formulation and solution of the pertinent optimization problems and increase the computational cost of finding a solution. Although full reservoir simulation provides detailed analysis and prediction of reservoir performance, the significant uncertainty and complexity of reservoir models can make the simulation results and their interpretations questionable. Moreover, in some cases, a reservoir model may not even be available to perform full simulation for performance optimization. The cost and complexity of developing full-scale simulation models, together with the considerable computational overhead associated with production optimization (especially under geologic uncertainty), call for development of fast proxy models for production optimization. To this end, various reduced-order and surrogate models have been designed to approximate the production behavior of a reservoir at a fraction of the computation required for full simulation.
We present an efficient production optimization scheme by integrating constrained optimization with fast decline curve analysis for predicting well production performance. The proposed production optimization approach is formulated as a constrained optimization problem by defining a desired objective function and a set of existing field/facility constraints. An efficient gradient-based optimization algorithm is then adopted to solve the resulting optimization problem for a single timestep. The optimization is then coupled with the decline curve analysis to predict future production rates. The optimization process is performed recursively in time for a specified duration. The predictions with the decline curve analysis are reasonable so long as the operating conditions remain unchanged. Using field data, we demonstrate that the proposed formulation can provide fast solutions to large-scale production optimization problems. The results in this paper suggest that the developed technique can be applied to improve production performance and operation efficiency with a minimal computational cost when compared to production optimization with full-scale reservoir simulation. It also offers the flexibility to adjust the problem formulation under various field conditions and is particularly useful when a full-scale reservoir model does not exist simulate the reservoir response for production optimization.