A Hybrid Artificial Intelligence Method for the Optimization of Integrated Gas Production System
- Hui-june Park (Seoul National University) | Jong-Se Lim (Korea Maritime U.) | Joo Myung Kang (Seoul National University) | Jeongyong Roh (Korea National Oil Corporation) | Bae-hyun Min (Seoul National University)
- Document ID
- Society of Petroleum Engineers
- SPE Asia Pacific Oil & Gas Conference and Exhibition, 11-13 September, Adelaide, Australia
- Publication Date
- Document Type
- Conference Paper
- 2006. Society of Petroleum Engineers
- 4.1.2 Separation and Treating, 4.6 Natural Gas, 4.1.5 Processing Equipment, 7.6.6 Artificial Intelligence, 6.1.5 Human Resources, Competence and Training, 4.1.4 Gas Processing, 5.1.5 Geologic Modeling, 5.5 Reservoir Simulation, 5.6.8 Well Performance Monitoring, Inflow Performance, 5.6.3 Deterministic Methods, 5.3.2 Multiphase Flow, 4.3.4 Scale
- 2 in the last 30 days
- 337 since 2007
- Show more detail
- View rights & permissions
|SPE Member Price:||USD 8.50|
|SPE Non-Member Price:||USD 25.00|
The optimization of gas production system has been largely based on a nodal analysis approach of individual wells. This approach limits the overall benefit of a field study because it ignores complicated flow interactions among wells during the optimization process, often resulting suboptimal operations. A major impediment in the formal optimization of large petroleum producing fields is the excessive computation time during optimizing processes. The conventional deterministic algorithm can not take into consideration the uncertainties regarding nonlinear and complex systems.
This paper proposes a new hybrid artificial intelligence approach for field-wide optimization of integrated production system including gas reservoir, well flow, and surface production network. An advanced polynomial neural network (PNN) with layer over-passing structure has been developed to replace a relatively time consuming reservoir simulator through robust and systematic search algorithm. The networks are subject to some form of training based on a representative sample of simulations that can be used as a re-useable knowledge base of information for addressing many different management questions.
For an optimal design of gas production systems, this study uses an integrated simulation model implemented by coupling a PNN with surface production network and optimization scheme based on fuzzy nonlinear programming approach to accommodate uncertainties by combining the fuzzy ?-formulation with a co-evolutionary genetic algorithm.
The proposed approach significantly reduces computational effort for optimization of the development scheme within reasonable accuracy. The hybrid optimization method can find a globally compromise solution and offer a new alternative with significant improvement over the existing conventional techniques. This technique provides optimum production rates of each well of the entire production system to deliver the target rates over the period of the gas contract requirements.This system has been validated by synthetic case study offering an important element for management of gas production strategies and field development.
The objective of production optimization is to determine operational parameters subject to all constraints at a given production period. Formal optimization strategies normally evaluate hundreds or even thousands of scenarios in search for the optimal solution to a given management problems1. The application of the optimization technique using a reservoir simulator as the forecasting tool is extremely time consuming and expensive. Fujii and Horne2 and Wang et al.3 have used local inflow performance relationships or material balance models to reduce the intensity of computationally demanding conventional methods. These approaches, however, were limited in ability to handle flow interactions among wells. Recently, artificial neural networks (ANNs) have been successfully used to solve many of these complex problems in optimization of field operations1,4. However, neural network training with extensive data still remains to be time-consuming. The back-propagation training algorithm with a gradient decent approach suffers from local minima problem, resulting in the production of unstable and non-convergent solutions.
This paper proposes a new hybrid model of PNN and genetic algorithm (GA) to replacethe original simulator for predicting reservoir performance. In this method, several key well scenarios selected by engineering judgment and/or randomly are evaluated using a numerical reservoir simulator. The simulation results form the basis for training and the trained network is used to make predictions of unseen production scenarios3. Thousands of possible scenarios are evaluated using the PNN with an insignificant computational effort maintaining reasonable accuracy.
|File Size||216 KB||Number of Pages||9|