| Popis: |
In the greenfield development process, one of the key questions that needs to be answered is, "What is the range of EUR for a particular development concept and the associated completion method based on the existing range of subsurface uncertainties?" The key challenge then is how can the team forecast a representative range of EUR efficiently to obtain a range of results that represent a probabilistic outcome. During the reservoir modelling process of this case study, a total of 405 static realizations had been run and then a STOIIP S-curve was generated. In the next step, 20 cases each of "High, Mid and Low" static models were selected based on the S-curve distribution for the next phase of dynamic simulation due to time and resources constraint. In terms of completion, the same development concept and completion method is assumed, where each dynamic case requires 8 horizontal producing wells with 200 metres of completion interval. Wells placement aside, each of the 60 dynamic models should not have the same fixed perforation depths and intervals due to the geological uncertainties with regards to facies distribution and they need to be selected based on the well effective k-h and hydrocarbon saturation along each well trajectory. Manual work could be used to analyse the best intervals for each of the planned wells, or in this case, this laborious process was replaced with an automated selection of the optimum completion interval workflow using Python script. This paper will show the workflow of how a scripted Python code is designed to provide an "automated moving window" to find the best intervals along a well trajectory. This workflow was executed in the pre-processor of the dynamic simulator which has a workflow window with Python-embedded capability. The Python code then generated the simulation keyword COMPDATMD, which contained the best perforation intervals for all the wells as an output. This automated workflow resulted in an optimization of the completion intervals in all the 60 dynamic model cases, while the ultimate recovery for this greenfield development in Peninsula Malaysia increased by 30% compared to EUR from previously "unoptimized runs". This approach is managed to cut down the run preparation time by at least two weeks compared to the manual solution. The improved range of EUR is also considered as a more representative outcome of the field development evaluation. Utilizing emerging technology breakthrough such as ability to customize specific features via a programming language is important towards a successful era of the Fourth Industrial Revolution (4IR). The results of this automated and customized workflow automation demonstrate a successful application of using machine learning for enhanced problem-solving in reservoir simulation. |
