CO₂ flooding in low-permeability and tight oil reservoirs is frequently compromised by severe gas channeling, which significantly reduces oil recovery and sweep efficiency. While foam flooding can effectively mitigate CO₂ channeling by trapping CO₂ within lamellae, its stability deteriorates under harsh high-temperature, high-salinity reservoir conditions, compromising its effectiveness. Furthermore, foam flow in porous media involves constant foam generation, collapse and propagation, making its flow behaviors difficult to predict. To address these challenges, a new foaming agent with satisfactory regenerative capability is developed to maintain gas mobility control under harsh reservoir conditions. The multiphase flow behaviors during CO₂ foam flooding are predicted using pore network modeling to obtain the corresponding relative permeability curves, which are further incorporated into a reservoir simulator to evaluate field-scale foam flooding performances, as well as optimize injection strategies. This multi-scale modeling approach establishes a quantitative link between pore-scale foam behaviors and field-scale oil recovery performances, offering new insights into carbon capture and utilization with enhanced oil recovery in low-permeability and tight reservoirs.

With the ongoing rise in global energy demand, the importance of enhanced oil recovery in oilfield development is becoming increasingly prominent. However, traditional chemical flooding agents face bottlenecks such as poor adaptability to application environments, unclear coupling mechanisms regarding multiple factors, as well as long research and development cycles. This paper systematically discusses the innovative paradigm of oilfield chemical agent development driven by artificial intelligence and proposes four core technological breakthroughs. Firstly, artificial intelligence-empowered molecular simulation technology can reveal the behavior mechanisms of flooding agents under extreme conditions. Secondly, intelligent molecular design using generative adversarial networks and reinforcement learning breaks through the traditional trial-and-error model. Thirdly, the construction of a data-mechanism dual-driven multi-objective optimization model achieves the collaborative prediction of physicochemical properties, economic benefits and environmental friendliness. Lastly, an integrated system of robotic chemist and high-throughput experimental platforms forms a closed-loop system of “artificial intelligence design - automated synthesis - online detection”, yielding a complete ecosystem. The analysis of the current technological development challenges and future development directions reveals that the artificial intelligence-empowered intelligent Research and Development system is expected to promote the transformation of chemical flooding technology toward efficiency, environmental protection and sustainable development, providing a new standard for intelligent oil and gas field development.