Application of hybrid numerical reservoir simulation and artificial neural network for evaluating reservoir performance under waterflooding

In the current study, an artificial neural network (ANN) and a numerical reservoir simulation (NRS) technique are used to analyse reservoir performance under waterflooding in the ZH86 block of the Zhaozhouqiao oilfield, China. Using five input datasets extracted from the history-matched NRS model, an NRS-ANN hybrid is trained using a trial-and-error approach. NRS-ANN hybrid model #46 (which has 5, 10, 10, 6, 6, and 1 neurons in the input layer, four hidden layers, and output layer, respectively) is found to produce the minimal root mean square error on the test dataset. On the validation data, the prediction performance of the selected NRS-ANN hybrid model achieves a minimal root mean square error of 0.0274 m³/day and maximal coefficient of determination and coefficient of correlation values of about 0.9999. The correlation between the block liquid production rate (BLPR, m³/day), block water production rate (BWPR, m³/day), block water cut (BWCT, %), block water injection rate (BWIR, m³/day), and block reservoir pressure (BRP, bar) as input variables and the simulated oil production rate (SOPRH) as the output variable is investigated. There is a positive correlation between SOPRH and BLPR, BWIR, and BWCT, and a negative correlation between SOPRH and BRP and BWPR. Segment B of ZH86 block experiences a 3.8% increase in BLPR, while segments A and C show declines of 1.3% and 1.6%, respectively. These variations in the liquid production rate correspond to changes in SOPRH of 4.3%, 1.9%, and 9.7% for segments A, B, and C, respectively. The prediction performance of the NRS-ANN hybrid model is compared with that of a simple NRS model. The accuracy of the NRS-ANN hybrid model in predicting oil production is found to be 1125 times that of the NRS model. Based on these results, it is concluded that the proposed NRS-ANN hybrid provides an accurate and useful tool for analysing reservoir performance under the waterflooding oil recovery technique.