Integrated Thermodynamic Reservoir Modeling through Efficient Design of Experiments for Optimal Field Development through Steamflooding Processes

摘要

The Steamflooding has been considered in this research to extract the discontinuous bitumen layers that are located at the oil-water contact for the heterogeneous light oil reservoir to improve oil recovery. The reservoir heterogeneity and the bitumen layers impede the water approaching into the reservoir from the infinite active aquifer; therefore, the Steamflooding has improved its efficiency to handle this situation. This research focused on using design of experiments (DoE) with thermodynamic reservoir flow modeling for the purpose of identifying the most sensitive factors that impact the reservoir performance through Steamflooding processes. Furthermore, the DoE helps to obtain the most likely scenario that achieve optimal reservoir response through the Steamflooding process. Meanwhile, the thermodynamic simulation modeling was used to evaluate the various what-if scenarios and compute the cumulative oil production that has been considered as a response in the experimental design procedure. In this paper, the conventional designed of experiments method has never applied; however, a new method of decimal points sampling has been adopted to handle more than three levels for each factor. The new method, which provides an efficient low-discrepancy sampling, compiles the Enhanced Stochastic Evolutionary (ESE) algorithm for Hammersley Sequence optimization to improve its discrepancy and uniformity. Because of its low discrepancy, Enhanced Stochastic Evolutionary (ESE) algorithm for Hammersley Sequence Sampling (ESEHSS) tends to sample space ”more uniformly” than the Hammer- sley Sequence itself and other random number sequences. The factors that have been considered to test the reservoir response are steam injection pressure, steam quality, steam injection rate, steam temperature, and number of injectors. To validate the overall design and each factor, t-distribution test and analysis of variance (ANOVA) test have been adopted in this study for modeling the relationship between the response and the factors by computing the sum of squared error and the variance of each factor. The stepwise elimination has been adopted to justify the DoE model to get the reduced linear model that represents most accurate simulation for the problem. The factors that have been identified by ESEHSS, considering the normal score transformation to get Gaussian distribution to the flow, response as most sensitive are steam quality, and some of the interaction terms that include other factors. However, the linear model of the original response has shown all the factors and interaction terms have an effect on the flow response.