Optimize Well Placement Based on Genetic Algorithm and Productivity Potential Maps

Abstract

Determining the optimal well location is a challenging task because the effects of geological and engineering variables on reservoir performance are often highly nonlinear and multimodal. The computational requirements for this problem based on automatic optimization are extensive, as many functional evaluations are required, each of which requires a complete reservoir simulation. Therefore, reducing the optimization time and improving the optimization effect is the key to promote the wide application of automatic optimization technology. In this study, we present a technique that combines the genetic algorithm (GA) with the helper method, productivity potential maps (PPMs) (GA + PPMs), to improve the effect of well placement optimization. The PPMs are generated by three typical methods: analysis method, numerical simulation method, and fuzzy system method. Numerical tests are carried out on three well placement methods in the PUNQ-S3 oilfield, namely, the original well placement and well placement proposed by GA and GA + PPMs plans. The result shows that generating the PPMs by an analytical method is the best choice. The cumulative oil production (COP) generated by GA + PPMs increased by 20.95% and 8.09%, respectively, compared with the original well scheme and GA well scheme, which demonstrates that the initial well location determined by reservoir engineers based on the PPMs has a significant impact on GA performance. Overall, the combination of GA and productivity potential maps is promising for this challenging task.