Steam huff and puff injection is one of the thermal EOR methods in which steam is injected in a cyclical manner alternating with oil production. The cost and time inefficiency problem of reservoir simulation persists in the design of a steam huff and puff injection scheme. Building predictive proxy models is a suitable solution to deal with this issue. In this study, predictive models of the steam huff and puff injection method were developed using two machine learning algorithms, comprising conventional polynomial regression and an artificial neural network algorithm. Based on a one-well cylindrical synthetic reservoir model, 6043 experiment cases with 28 input parameter values were generated and simulated. Outputs from the results such as cumulative oil production, maximum oil production rate and oil rate at cycle end were extracted from each simulation case to build the predictive model. Reservoir properties that could change after an injection cycle were also modeled. The developed models were evaluated based on the fitting performance from the R-square value, the mean absolute error (MAE) value and the root mean square error (RMSE) value. Then, Sobol analysis was conducted to determine the significance of each parameter in the model. The results show that neural network models have better performance compared to the polynomial regression models. Neural network models have an average R-square value of over 0.9 and lower MAE and RMSE values than the polynomial regression model. The result of applying the Sobol analysis also indicates that initial reservoir water saturation and oil viscosity are the most important parameters for predicting reservoir production performance.

Application of predictive model in enhanced oil recovery has been mainly to obtain an estimate of production performance. The advantage of a predictive model over reservoir simulation are the computational speed and its significantly lower cost. A predictive model is able to provide results instantly, comparing to the whole process of reservoir simulation. However, it should be noted that predictive model is not a substitute for reservoir simulation, but rather as a starting approach for field development planning purposes. Hydrocarbon injection is an EOR method which incorporates the injection of hydrocarbon gas to increase oil production. Presently, there has been little to none predictive models developed to estimate oil production for hydrocarbon injection method. In this study, a predictive model of the hydrocarbon injection method is presented. The predictive model was developed using commercial software CMG. Based on an inverted 5-spot reservoir model, thousands of reservoir simulation cases with different parameter values were conducted as a sensitivity analysis. Cumulative oil production result was captured from each simulation case to build the predictive model. Polynomial regression and neural network predictive models were built. The neural network model fitted the simulation data better than the regression model. R-square value for both predictive models exceeded 90%. This predictive model can be confidently used to estimate cumulative oil production as long as the input parameter values are within the parameter intervals of the model.

No abstract

CO2 Injection is one of the EOR methods that holds potential to be widely implemented. However, the underlying issues in implementing CO2 injection include the availability of CO2 sources as well as the steep capital cost associated with the surface facilities required to deliver and process the CO2. In such case, the cyclic injection, so-called Huff & Puff CO2 injection is seemingly more feasible. In this study, we propose an alternative predictive model to determine the performance of CO2 cyclic injection. This model can be utilized as a preliminary screening tool and to understand the feasibility of the operation in a quick manner. In addition, the uncertainties associated with the reservoir conditions can also be considered within the model. Most of the efforts to develop predictive model for CO2 cyclic injection have been concentrated around empirical approach, and hence, only limited parameters are considered. In this study, we establish experimental design through reservoir simulation and perform machine learning to construct the predictive model. Therefore, most important factors driving the performance of the well underwent CO2 cyclic injection including reservoir rock and fluid properties, interaction between rocks and fluids, drainage geometry, and operating condition are incorporated. Compared to the simulation results, the proposed predictive model shows an acceptable accuracy with R-squared of above 0.95. The result of this study can be used to help find suitable well and reservoir to be injected with cyclic CO2 injection, as preliminary analysis on the individual well performance can be conducted before performing full field simulation. The model can also be utilized to determine the optimum amount of CO2 needed, as well as the length of injection and soaking time before the well is put into production.