Auto-tune of PVT data using an efficient engineering method: Application of sensitivity and optimization analyses

In reservoir simulation, it is crucial to have a physical model of the reservoir fluid that can give the properties and behavior of the fluid at different conditions. Phase behavior of a reservoir hydrocarbon is modeled through an equation of state (EOS) by adjustment of some of the parameters in EOS so as to reproduce measured data of the mixture. Since the EOSs have some inherent deficiencies which may cause erroneous predictions, these models have to be tuned against experimental data. Conventionally, the methods and algorithms that have been proposed for tuning EOS models are tedious and really time consuming; therefore, a simple and accurate algorithm that can find the best match between EOS and measured data, automatically is really needed. For this purpose, a generalized auto-tune procedure based on Monte-Carlo and genetic algorithms is proposed in this study. In the presented method, first of all, the matching parameters for regression process are found in an automatic manner by using Monte-Carlo sensitivity analysis. Then, the selected parameters are fed to the PVT analyzer coupled with genetic algorithm and therefore, the best EOS model is searched and found. Three fluid samples (one gas condensate and two black oils) were used in order to test the proposed method and the matches were found by using Peng-Robindon (PR) EOS. The results illustrated the surprising high performance of the developed algorithm in matching the experimental data. The main advantages of this method is its high speed in finding the best EOS model, working automatically, reducing costs and also having the possibility of evaluating the model at different conditions.