A sequential well-placement strategy is important for field-development planning under geological uncertainty because reservoir conditions evolve between drilling stages. This study proposes a hybrid framework that combines convolutional neural networks (CNNs) and a genetic algorithm (GA) to optimize well locations while reducing the number of full-physics reservoir simulations required during optimization. Open Porous Media (OPM) Flow was used to generate training datasets for two consecutive six-month periods, allowing the CNN proxies to learn the relationship between permeability realizations, well coordinates, and cumulative oil production (COP). The trained proxies were then coupled with the GA to perform sequential well-placement optimization. The methodology was evaluated using the Egg reservoir model with three selected permeability realizations. The CNN proxies achieved coefficients of determination of 0.84 and 0.81 for the training subsets and 0.71 and 0.74 for the testing subsets in Periods I and II, respectively. The corresponding mean absolute percentage errors were 2.3% and 1.3% for the testing subsets. After validation with OPM Flow, the proxy-derived optimal solutions showed relative differences below 1% in ensemble-mean COP compared with the reservoir simulator. The proxy-based workflow required about 26% of the computational time of direct optimization in Period I and about 15% in Period II. The main contribution of this work is a sequential CNN–GA workflow in which separate proxy models are trained for successive drilling periods, allowing well-placement decisions to account for the evolved reservoir state between development stages.

A sequential well placement strategy is important for field development planning under geological uncertainty, because reservoir conditions can change between drilling stages. Motivated by this challenge, this study proposes a hybrid framework that combines convolutional neural networks (CNNs) with a genetic algorithm (GA). The goal is to determine optimal well locations efficiently, while reducing reliance on full-physics reservoir simulation. The methodology uses OPM Flow to generate training datasets for two consecutive six-month periods. This allows the CNN proxy to learn the relationship between permeability realizations, well coordinates, and cumulative oil production. The trained proxies then guide the GA-based optimization in each period. Results for the Egg model show strong predictive performance in both stages. The coefficients of determination are 0.76 and 0.82 for training data, and 0.64 and 0.63 for testing data, in the first and second periods, respectively. In addition, the proxy-based optimization required only about 26% of the computational time of direct simulation in the first period, and roughly 15% in the second. Production estimates were maintained within a 5% error margin. Overall, the proposed sequential, proxy-assisted approach is accurate and computationally efficient for well placement optimization under geological uncertainty.