Simulating and Optimizing Production from a Multi-Well Field
MetadataVis full innførsel
In this thesis a system for automation and optimization of a PIPESIM production network model has been developed, building on a system that a previous master student made. The system uses Pipe-It, a program developed by Petrostreamz AS, and a combination of VBScript scripts and Excel VBA macros. By using these scripts and macros the Pipe-It system can interact with PIPESIM through Open Link, an application programming interface (API). The system can adjust ESP (electric submersible pump) and PCP (progressive cavity pump) speeds and turn wells on and off in the model, before running the simulation of the model in PIPESIM. After the simulation the system gets various results from different PIPESIM result files. Pipe-It automates the running of the different scripts and macros, and can thereby be used to optimize the PIPESIM model by adjusting pump speeds and turning wells on and off. The Pipe-It system also has different parts for comparison and analysis of results, and a small part for simple economic calculations. The system has an automated optimization setup as well, which can be used in this case to set up the optimization much quicker than by using the built-in Optimizer in Pipe-It. In addition the system can extract data and settings from some specifically formatted field reports, transfer those settings to the model, and compare field data with simulation results.A combined optimization including both continuous variables, pump speeds, and integer variables, turning wells on or off, has been tested in Pipe-It using a simplified PIPESIM model. The optimization did not manage to find the global maximum for total oil production, but in the best case it did manage to find the optimum combination of wells to turn on and off. In all cases the total oil production was improved compared to the initial starting point and was close to the global maximum. But the optimization required a long run time, ranging from around 14 hours to over 60 hours. A nearly identical model was also implemented in Excel and the combined optimization tested using the built-in Excel Solver. The Excel Solver was also not able to find the global maximum, and based on previous experience the results will usually be better when using Pipe-It with a PIPESIM model.A simplified case of only turning wells on and off with pump speeds constant was also tested, both in Excel and using Pipe-It. In this case both the Excel Solver and the Optimizer in Pipe-It was able to find the optimum combination for the wells.As the model tested was small and simplified, a larger and more realistic model is expected to take much longer time to optimize when changing pump speeds and turning wells on and off. Based on that and earlier experience, it is recommended to instead perform optimization with only pump speed being adjusted. Wells can then be turned off manually based on analysis of the optimization results and any field data and experience, before optimizing the model again if necessary. This could possibly find a global optimum, or atleast an improved solution, a lot faster, as optimization only involving adjustments to pump speeds will require less time.