When processing the retrieved multiphase fluid from the wells, the many types of equipment used consume a great amount of energy during the separation of oil, gas and water. Specific oil and gas export quality has to be met before the products can be exported to the consumer. Achieving the export quality and minimising the power consumption is a complex system to predict.
This projects deal with an approach were Design of Experiment and Response Surface Methodology has been used in order to create a regression model of a given HYSYS separation simulation, with five independent high influence variables, which can predict a global minimum with export constraints applied. An optimisation algorithm was created in the programming language Python and it was verified to perform well inside the experimental training set. The optimisation model was found suitable to future industry usage, in order to find the lowest possible power consumption and still maintaining the strictly demanded export quality constraints.
In the extension of the optimisation model, the HYSYS separation simulation was used to determine a regression model, estimating the central process equipment used, based on the variety of the flow rates for oil, gas and water. This set the foundation for a Monte Carlo model, which predict the total topside weight by using correlation factors. All contributions in the Monte Carlo model are independent randomly picked, within a normal distribution uncertainty applied to all aspects in the model. The model is intended to be used in future early phase projects for a less time consuming and more precise result, than the methods used at the moment.