Data-Driven Approaches for Quick Performance Assessment of Hydrocarbon Reservoirs

Hydrocarbon reservoir's heterogeneity, production, injection and aquifer support play a unique role into how field strategies are developed knowing key reservoir properties (e.g. drainage volumes, reservoir energy, rock properties, decline analysis, etc.). Alanood et al. (2022) illustrated applications of Fast Marching Method (FMM) in assessing reservoir performance, identifying reservoir patterns and anomalies from production/injection data, and predicting the reservoir response when considering modeling uncertainty for model calibration. The objective of this paper is to extend the previous work to include more field development controls like voidage replacement, infinite acting reservoirs, and performance prediction of new proposed producers or injectors. The proposed approach in this work is to extend the physics-aware data-driven approach, that uses the diffusive time-of-flight (DTOF) to determine the pressure disturbance caused by production/injection wells to help in comprehending the impact on field development. The previous work by Alanood et al. (2022) is extended to account for voidage replacement, pressure maintenance and infinite acting reservoirs. In addition, this approach has been extended to predict the performance (e.g., rates, bottom-hole pressure and plateau) of proposed injection and production wells to give a quick insight about their performance prior to running the full-physics reservoir simulators. Thus, will give a better understanding in describing the transient flow. That is in addition to helping in inferring faults existence, fracture networks and inter-well connectivity. In this work, SPE10 and NORNE benchmark models were used to test and validate the approach. The results show that this date-driven approach was able to match historical and predicted performance of the reservoirs and wells. The data-driven algorithms were able to generate pressure maps resembling the ones obtained from the full-physics reservoir for infinite acting reservoir models for the first few years before it starts to deviate. They also were able to predict rates and flowing bottom-hole pressure of injectors and producers to up to 80% when compared to reservoir simulation. In addition, several blind tests of placing injectors and producers revealed that the algorithms are capable of capturing the reservoir's pressure changes and wells performance in finite and infinite acting reservoirs. The extended work of the physics-aware data-driven approach using the DTOF is a reliable approach to monitor reservoir pressure in the finite and infinite acting reservoirs (e.g., aquifer support and pressure maintenance injection). It also demonstrated the ability to predict the performance of existing and newly proposed wells to great accuracy. The approach provides a quick assessment of wells and reservoir performance given a certain field development strategy.