Optimal Field Development and Control Yields Accelerated, More Reliable, Production: A Case Study Morteza Haghighat, Khafiz Muradov, David Davies Institute of Petroleum Engineering, Heriot-Watt University, Edinburgh, UK SPE Aberdeen 3rd Inwell Flow Surveillance and Control Seminar 3 October 2017
2/18 Outline Introduction to Intelligent Wells (I-wells) I-wells control Reactive Proactive Challenges in Proactive Control Developed framework for proactive optimisation under reservoir description uncertainties Intelligent field development case study Conclusions Extensions Acknowledgements
3/18 Introduction Intelligent Wells (I-wells) Equipped with down-hole monitoring and control devices ICD (Inflow Control Devices) Single, fixed position AICD (Autonomous Inflow Control Device) self-adjusting position, providing a pre-designed, fluid-dependent flow control ICV (Interval Control Valve) Multiple positions, surface control ICV provides a flexible production control, BUT maximum Added Value depends on identifying the optimal ICV control strategy From SPE-107676 with modification
4/18 Introduction Reactive Control Strategy of ICVs Reactive Decisions are based on the system s current condition Considers Short-term (current) objectives Production Improvement Fast reaction to recognised situations Potentially can be done using well intervention
5/18 Introduction Proactive Control Strategy of ICVs Proactive Starts earlier Mitigates future undesired problems and/or states. Long-term objectives increased Oil Recovery Requires a reservoir model to forecast production
6/18 Proactive Optimisation Problem Formulation & Challenges Objective: Find the control scenario of ICVs that maximises the objective function ICVs Control Simulated Reservoir Model With Uncertainty Objective function (e.g. NPV) Challenges Large number of control variables Computationally expensive objective function evaluations (i.e. reservoir simulator) Uncertain objective function No. of control variables = No. ICVs Total Control period length of control steps
7/18 Proactive Optimisation Developed Robust Optimisation Framework A fast and efficient optimisation algorithm is developed which can handle large number of control variables with minimum obj. fun. evaluations SPE-167453, SPE-178918 Accounting for reservoir description uncertainty Optimiser ICVs Control Reservoir Model Reservoir Model Reservoir Model Modified Objective function - Mean optimisation: Search for Objective a control function scenario (e.g. which NPV) improve all realisations (to some extent)
8/18 Case Study Model Description & Development Plan A full-field, consists of two overlaying heterogeneous reservoirs, each divided into two layers 4 zones, 4 ICVs & 4 Packers to separate zones Conventional Development Plan: - 14 producers (single zone) - 7 injectors Alternative I-well development plan: - 3 intelligent producers (commingled) - 8 conventional producers (single zone) - 7 injectors
Normalized Field Oil Production Rate Normalized NPV 9/18 Case Study Reservoir Description Uncertainty Formation porosity and permeability, faults (locations and transmissibility), initial water saturation and reservoir net-to-gross were the major uncertainties. 3 realisations known as P10 (optimistic), P50 (base) and P90 (pessimistic) are employed to capture this uncertainty Enough to capture the underlying uncertainty?! 1.2 1 0.8 0.6 0.4 0.2 P10 P50 P90 0 01/00 09/02 06/05 03/08 12/10 Date (mm/yy) 1.2 1.0 0.8 0.6 0.4 0.2 0.0 01/00-0.2 09/02 06/05 03/08 12/10-0.4 P10 P50 P90 Date (mm/yy)
% Increase in NPV above the base case 10/18 Robust Proactive Optimisation Effort Total Optimisation time using 18 CPUs was lass than 1.5 days (~ 33 hr). More than 80% of the improvement was obtained after 10 iterations requiring only ~3.5 hours computation time. 5 4 Mean P10 P50 P90 The developed framework is capable of performing proactive optimisation of ICVs in a reasonable time for this relatively large, full-field model. 3 2 1 0-1 0 20 40 60 80 100 Iteration Number
% Change in mean and variance compared to an I- well with fully-open ICVs % incease in NPV compared to an I-well with fullyopen ICVs 11/18 Robust Proactive Optimisation Added-Value 4.0% 2.0% 0.0% -2.0% -4.0% -6.0% -8.0% -10.0% -12.0% -14.0% Mean Variance Mean and Variance of all realisations Improved mean higher expected added-value reduced variance higher reliability (lower risk) by applying the best ICV control scenario 5.0% 4.5% 4.0% 3.5% 3.0% 2.5% 2.0% 1.5% 1.0% 0.5% 0.0% P90 P50 P10 Reservoir Models greater improvement for the less favourable realisations An extended oil plateau was observed for all realisations.
NPV(10^8 $) ΔNPV w.r.t. conventional well (10^7 $) 12/18 Conventional Vs. I-well Development Plan Impact of Robust Proactive Optimisation Conventional Development plan 14 conventional producers I-well development plan 8 conventional & 3 intelligent producers 20 15 Extended plateau by robust proactive opt. I-well (Fully Open) I-well (Optimised) End of Plateau Period End of Drilling 25 10 20 15 10 5 0-5 -10 P90 realisation Conventional-Well I-well (Fully Open) I-Well (Optimised) Date 5 0 09/01 01/03 05/04 10/05 02/07 07/08 11/09-5 Lower number of wells drilled Date (mm/yy) Accelerated field development I-wells lead to loss mitigated by optimum control
13/18 Robust Vs. Single realisation Proactive Optimisation Proactive control must be applied early, during the plateau period, to achieve the highest gain the reservoir model is most uncertain during the early period. The importance of robust optimisation is shown by considering a non-robust optimisation performed using a single realisation (here P50). Higher Added-value % Change P10 P50 P90 Mean Variance Robust optimisation +0.2 +1.3 +4.3 +1.6-12 Reduced uncertainty Singlerealisation (P50) +0.05 +2 +0.3 +0.8 +2.8 Max improvement for P50 but non-optimum performance for other realisations, lower added value, increased uncertainty
14/18 Conclusions Reservoir-model-based proactive control should be applied early, during the plateau period greatest uncertainty in the model Single realisation optimisation sub-optimal performance, high risk. One of the main reasons why the operators are often unwilling to control the ICVs/wells proactively. Although a no-control scenario may diminish the I-well gain! Robust proactive optimisation is the solution. Developed robust optimisation framework can efficiently handle large number of control variables, high computation time and reservoir description uncertainty The whole process was performed using a single, high-end PC in a reasonable time for this relatively large, full-field model
15/18 Conclusions (Partial) I-well development scenario Increased, early-time, NPV by reducing the number of wells to be drilled May accelerates field development by speeding up the drilling process state-of-the-art, proactive optimisation extended the oil production plateau, ensuring that the early NPV gain was maintained. Robust proactive optimisation allows the production operators to confidently control their I-wells to achieve maximum expected added-value lower uncertainty in the operation
16/18 Extension Realisation selection algorithm Reservoir description uncertainty is quantified by hundreds of model realisations How to select a small ensemble 3 of realisations as the 2 representative of all realisations P10, P50 & P90 are not always good enough representatives Developed realisation selection algorithm: smartly select an ensemble of realisations. Tailored to the subsequent application A. Visualisation B. Clustering 1 0-1 -2-3 -3-2 -1 0 1 2 3 Each circle is one model realisation Haghighat Sefat, M., Elsheikh, A. H., Muradov, K. M. & Davies, D. R. 2016a. Reservoir uncertainty tolerant, proactive control of intelligent wells. Computational Geosciences, 20, 655-676.
17/18 Extension Robust Completion Design The developed robust optimisation framework can be extended to advanced completion design Control Variables (Completion design parameters) - Type of Flow Control Devices (FCDs): ICV, ICD, AICD(V) - Location, Number (& strength) of FCDs - Autonomous, fluid dependent performance of AFCDs Optimiser Reservoir Model Reservoir Model Reservoir Model Objective function (e.g. NPV, cumulative oil) To be presented in Inflow Control Technology (ICT) Forum, 12 th & 13 th October 2017, San Antonio, USA.
18/18 New Phase of Value from Advanced Wells JIP (2018-2021) Theme A: Maximum Added value from downhole flow control completions Theme B: In-well monitoring and data interpretation in advanced wells Modelling Design Control Analysis Interpretation Data mining AFCDs AFCD completions TIFs Robust Prod./Inj. with AFCDs considering - Uncertainties - TIF Robust ICV control - Large fields - Uncertainties PTTA DTS oil and gas wells Test design Missing data Other topics. Sponsor steered
19/18 Thanks For Your Attention. Morteza Haghighat Acknowledgements The authors are grateful to the sponsors of the Value from Advanced WElls (VAWE) Joint Industry Project at Heriot- Watt University for funding