Well placement optimization with a novel swarm intelligence optimization algorithm: Sparrow Search Algorithm

Well placement optimization plays a critical role in developing oil and gas reservoirs, presenting a formidable challenge due to the high dimension of the search space and computational requirements. This research addresses the simultaneous optimization of well location and flow rates for production and injection wells in a heterogeneous oil reservoir located in southwestern Iran. The objective function employed to evaluate the proposed solutions generated by the Sparrow Search Algorithm (SSA) is the net present value (NPV). To enhance the performance of SSA, a sensitivity analysis is conducted to tune its hyperparameters, while keeping the epochs and population size fixed to manage runtime effectively. The study tackles the challenges posed by reservoir heterogeneity and the time-consuming nature of the optimization process through various strategies. The decision variables are reduced to a reasonable order, and a quality index is introduced to guide the algorithm towards exploring areas of the reservoir with higher potential. Furthermore, runtime is utilized as a termination condition to optimize computational time. The inclusion of the quality map significantly improves the NPV outcomes, allowing for more effective well placement decisions. Physical constraints related to well placement are handled using a penalty method and map cleaning techniques. Having addressed these challenges, the study investigates the impact of fixed parameters on the optimization results. By conducting an optimization run with 100 epochs, comparable outcomes to the case of 50 epochs (with a 5% improvement) are achieved, albeit with a longer runtime of 20 h. However, increasing the population size substantially raises both runtime and computational costs. Therefore, runtime is considered a practical termination condition, and efficient runtimes of 10 and 15 h are selected for this specific problem. The results indicate that a population size of 25 sparrows outperforms the cases with 50 and 100 epochs, respectively, yielding slightly higher NPV values. Furthermore, the performance of SSA is compared to that of Particle Swarm Optimization (PSO) in terms of NPV, convergence, and runtime. The results highlight the advantages of SSA, demonstrating faster convergence and achieving higher NPV values compared to PSO. However, the computational cost of SSA should be taken into consideration, as it requires significantly more time to reach the optimal solution compared to PSO. Overall, this research emphasizes the high potential of the SSA algorithm for optimizing well placements and its practical relevance in real-world oil industry cases.