Inferring the interwell connectivity of multilayer waterflooded reservoirs accounting for incomplete injection/production profiles

Abstract

The capacitance–resistance model (CRM) has been widely used to identify the interwell connectivity in waterflooding reservoirs. Although this technique has been extended to deal with the blocks having multiple layers, almost all these models require that the injection profiles of all the injectors or the production profiles of all the producers should be available. This demanding requirement limits its applications. We propose a multilayer capacitance–resistance model (ML-CRM) to identify the interwell connectivity within each layer using incomplete injection/production profiles. All the known injection/production profiles are input into the model directly, whereas the unknown profiles of the other wells are estimated by matching the historical production data. Based on the long short-term memory (LSTM) network, we also propose a method to identify the connectivity accounting for well shut-in. Our model is solved by coupling the biogeography algorithm (BBO) and the least-squares regression. We examine the performance of ML-CRM using both the data generated from numerical simulations of different synthetic fields and actual production data from the Shengli oilfield, China. We also discuss the noise effect on the model performance. Our results suggest that even though limited injection/production profiles are available, our proposed model fully utilizes the known information and can still provide reliable results. A comparison with the interwell tracer test in the real field indicates that this model well captures the different connectivities within the two layers, which is beyond the capability of traditional CRM.