Well placement optimization under time-dependent uncertainty using an ensemble Kalman filter and a genetic algorithm

• A new algorithm for well placement under time-dependent uncertainty is presented.
• The algorithm combines ensemble Kalman filter and genetic algorithm.
• The application to PUNQ-S3 reservoir with 50 geological realizations is presented.
• Results show significant improvement in efficiency compared to the previous work.

Determining the optimal well location in a reservoir is a challenging problem. It involves taking several factors into account, including geological uncertainty, reservoir and fluid properties, economic costs, and technical ability. Most research on well placement optimization under uncertainty has assumed static uncertainty in the reservoir parameters, until the introduction of the pseudohistory concept. The pseudohistory concept incorporates the field's probable history and results in the determination of optimal locations of future wells with greater certainty. This approach, however, requires an excessive number of simulations and may not be practical for optimization of a reservoir model having a large number of geological realizations.In this study, we use an ensemble Kalman filter (EnKF) to perform history matching of the PUNQ-S3 reservoir model using data from six production wells over an eight-year period. This is followed by well placement optimization using a genetic algorithm (GA) combined with pseudohistory matching, carried out over two years, following the placement of the first future well. Thus, this approach not only provides increased certainty in optimal well placement but also, using EnKF as a history matching method, requires only a single “best estimate” realization for objective function evaluation during GA optimization. As a result, the total time taken to find the optimal well locations is significantly reduced. We illustrate this through comparison with the previous research.