Real-time fracture treatment optimization has been an aspiration for decades, with modest progress in conventional reservoir tip screen-out designs using rapid mini-frac analysis to adjust pad size and proppant schedules. Real-time treatment optimization in unconventional reservoirs, once thought to be impossible, is now being pursued by operators and service companies. Service companies are automating equipment and enabling "intelligent" completions with low-cost measurements, while operators are envisioning models that can use these low-cost measurements for real-time optimization. However, the value of real-time optimization has not been studied. This paper quantifies the likely value of the perfect frac stage.
The vision of Autonomous and Intelligent Fracturing (AIF) is to enable real-time or stage level improvements in treatment design and/or completion strategy. To realize the AIF vision, there are four major components that are currently missing: (1) fast optimization models, (2) low-cost measurement technologies, (3) reliable fracture geometry control technologies, and (4) the value proposition. This paper introduces the AIF vision and discusses ongoing work to develop fast optimization models and low-cost measurement technologies, but the focus is on the value proposition. Given the operational and subsurface complexities of pad-scale completions and the technology challenges of real-time optimization, we need to understand if such an ambitious goal as AIF is worth the cost. This study focuses on the value of the perfect fracture treatment stage, defined as achieving the design goal of equal fluid and proppant in every cluster.
Two modeling studies were conducted to estimate the value of the perfect frac stage, one using a fully coupled hydraulic fracture-reservoir simulation model and a second study using a simple fracture-reservoir simulation model. The fully coupled model was calibrated using extensive field measurements in the Bakken, including DAS measurements of cluster-level fluid distribution, strain measurements of fracture morphology, fracture propagation pressures from observation lateral gauges, and microseismic measurements of fracture geometry. Both models included five wells and varied well spacing from 440 ft to 1100 ft. Cluster-level fluid distribution was varied from poor to an expected base-case to perfect. The results suggest that well productivity can be improved by as much as 20% if fluid distribution is poor and the perfect stage can be achieved. However, cluster-level fluid distribution may not be "poor" and the perfect frac stage may not be attainable. A conservative estimate for the productivity increases with real-time optimization is discussed in the paper using actual DAS uniformity measurements.