DATA DRIVEN MAPPING OF LIQUID SATURATION IN WATERFLOODS USING INJECTION AND PRODUCTION DATA

Mapping spatial distribution of oil and water saturations in a waterflood can help in recognizing sweet‑spots for infill wells. Common methods of reservoir simulation and pressure pulse techniques rely on pressure and rate data for interwell transmissibilities estimation and storativities leading to indirect estimation of saturation. We propose an innovative data‑driven approach that enables categorization of liquid saturation in waterfloods solely based on injection and production data. Particularly, our approach enables 1)prediction of the gross production rates of the producers and 2)mapping of the average liquid saturation among injector‑producers in a waterflood based on the predicted production rates. Accordingly, we have implemented our approach as a two‑step process. First, we derive the injection allocation factors to predict the production rates at the producers given the injection rate schedule. Without loss of generality, we assume a linear relationship between production and injection rates which is modeled using Multiple Linear Regression analysis. Next, we form and solve a system of equations based on the transmissibility equations to categorize nature of liquid saturation between injector‑producer pairs. Areas where water saturation is high are detected from estimated tranmissibilities derived from allocation factors. Finally, we need to categorize and map water saturation distribution between injector‑producers across the waterflood. To this end, we first normalize the transmissibility values among wells.We then categorize the saturation regimes from the normalized values and develop areal distribution of sweet‑spots by indicator kriging of the estimated oil saturation categories. Consequently,our approach offers a low‑cost, fast and scalable technique without requiring reservoir simulation. We have successfully tested our approach with synthetic waterfloods with two different patterns (five‑spot and linear) in various scales. We have verified our results by comparing saturation categories with those predicted by the commercial simulator.