## Abstract

Intelligent Wells are distinguished from conventional wells by being equipped with downhole sensors to monitor the Inflow Control Valves (ICVs) to control the (multiple) zonal flow rates. The data from the downhole sensors monitors the properties of the fluid flowing into the well from the reservoir at a zonal or a well level. The sensor data is analysed to provide the necessary information for the ICVs to be operated in the optimum manner i.e. to increase the hydrocarbon recovery and prevent unwanted fluid production.

This objective is simply stated, but the optimisation calculations required to identify the optimum ICV settings necessitates the repetitive solution of a complex, non-linear problem. Several commercial software providers have made such optimisation algorithms available to the industry to perform this task. However, experience has shown that challenges still arise when they are applied to large, complex models even though these algorithms work well on many simple cases. This is especially true when the optimisation algorithm is combined with a large, multi-well simulation model of multiple reservoirs with a complex, surface production network that is typical of those used today by operators to study real-field cases prior to field development.

Inclusion of the optimisation algorithm not only dramatically increases the calculation time (up to 50 times when compared with the equivalent run without such optimisation); but also stability and convergence problems give additional increases in the running time. More importantly, the combined software will sometimes simply stop, due to erroneous control parameters being provided by the optimisation algorithm. The optimisation algorithm may also return unrealistic results at random time intervals, a problem that can lead to unnecessary complications as it may not be immediately recognised. Such problems are particularly acute if the software is performing multiple realisations, for example when it is being applied to analyse the impact of a multiple field development scenarios or when studying how uncertainty in the reservoir’s dynamic and static properties affect the field’s production performance.

This paper will present a novel method based on the direct search algorithm for implementing an ICV control strategy. This method was chosen since it is not affected by the convergence problems which have caused many of the difficulties associated with previous efforts to solve our non-linear optimisation problem. Our control strategy will use the current, zonal inflow rate and water cut data to identify the optimal ICV choke positions. The availability of this data reduces the number of possible choke positions that have to be evaluated at each time step by the simulator. Run times similar to the base case are potentially possible while, equally importantly, the optimal value identified is similar to the value returned by the other published optimisation methods referred to above.

This paper outlines the assumptions made and, after exploring the method’s use in two single well models for reactive control of oil production from intelligent wells completed with discrete ICVs, its application to a large, reservoir simulation model will be illustrated. The latter application could be implemented rapidly, unlike some other optimisation software, because “tuning” of the model and/or the method was not required; the control algorithm being always convergent, fast and stable.

The proposed approach is particularly valuable for the analysis of the impact of uncertainty of the reservoir’s dynamic a static parameters. This arises because the modified direct search method employed here, being convergent and independent of the initial point, ensures that the result from the multiple realisations are directly comparable because “tuning” of the algorithm’s parameters are not required in the middle of the calculation procedure.

This objective is simply stated, but the optimisation calculations required to identify the optimum ICV settings necessitates the repetitive solution of a complex, non-linear problem. Several commercial software providers have made such optimisation algorithms available to the industry to perform this task. However, experience has shown that challenges still arise when they are applied to large, complex models even though these algorithms work well on many simple cases. This is especially true when the optimisation algorithm is combined with a large, multi-well simulation model of multiple reservoirs with a complex, surface production network that is typical of those used today by operators to study real-field cases prior to field development.

Inclusion of the optimisation algorithm not only dramatically increases the calculation time (up to 50 times when compared with the equivalent run without such optimisation); but also stability and convergence problems give additional increases in the running time. More importantly, the combined software will sometimes simply stop, due to erroneous control parameters being provided by the optimisation algorithm. The optimisation algorithm may also return unrealistic results at random time intervals, a problem that can lead to unnecessary complications as it may not be immediately recognised. Such problems are particularly acute if the software is performing multiple realisations, for example when it is being applied to analyse the impact of a multiple field development scenarios or when studying how uncertainty in the reservoir’s dynamic and static properties affect the field’s production performance.

This paper will present a novel method based on the direct search algorithm for implementing an ICV control strategy. This method was chosen since it is not affected by the convergence problems which have caused many of the difficulties associated with previous efforts to solve our non-linear optimisation problem. Our control strategy will use the current, zonal inflow rate and water cut data to identify the optimal ICV choke positions. The availability of this data reduces the number of possible choke positions that have to be evaluated at each time step by the simulator. Run times similar to the base case are potentially possible while, equally importantly, the optimal value identified is similar to the value returned by the other published optimisation methods referred to above.

This paper outlines the assumptions made and, after exploring the method’s use in two single well models for reactive control of oil production from intelligent wells completed with discrete ICVs, its application to a large, reservoir simulation model will be illustrated. The latter application could be implemented rapidly, unlike some other optimisation software, because “tuning” of the model and/or the method was not required; the control algorithm being always convergent, fast and stable.

The proposed approach is particularly valuable for the analysis of the impact of uncertainty of the reservoir’s dynamic a static parameters. This arises because the modified direct search method employed here, being convergent and independent of the initial point, ensures that the result from the multiple realisations are directly comparable because “tuning” of the algorithm’s parameters are not required in the middle of the calculation procedure.

Original language | English |
---|---|

Pages | 1-13 |

Number of pages | 13 |

DOIs | |

Publication status | Published - Jun 2012 |

Event | SPE Europec/EAGE Annual Conference - Copenhagen, Denmark Duration: 4 Jun 2012 → 7 Jun 2012 |

### Conference

Conference | SPE Europec/EAGE Annual Conference |
---|---|

Country/Territory | Denmark |

City | Copenhagen |

Period | 4/06/12 → 7/06/12 |