Study on seepage characteristics and dynamic optimization algorithm of reservoir during water polymer flooding

The efficiency of traditional water flooding technology in reservoirs with low permeability, strong heterogeneity or high viscosity decreases significantly, while polymer flooding, as an important branch of chemical flooding technology, can effectively improve the displacement effect by increasing the viscosity of displacement fluid and reducing the mobility ratio of water to oil. However, in the process of water polymer flooding, the reservoir seepage characteristics are complex, involving multiple physical and chemical processes such as multiphase flow, non-Newtonian fluid behavior, polymer adsorption retention and shear degradation, and the traditional static optimization method is difficult to adapt to the dynamically changing reservoir conditions. In this paper, a multi-scale seepage model considering polymer adsorption, shear degradation and non-Newtonian fluid behavior is constructed, and the accuracy of the model is verified by core displacement experiments and numerical simulation. On this basis, a hybrid intelligent optimization algorithm combining particle swarm optimization (PSO) and deep reinforcement learning (DRL) is proposed to realize efficient real-time regulation and optimization of injection parameters. The experimental and numerical simulation results show that the hybrid optimization algorithm not only significantly improves the oil recovery, but also reduces the final pressure difference and polymer retention, and is superior to the traditional PSO and pure DRL methods in convergence speed and economic benefits. This study provides a new theoretical basis and technical support for water polymer flooding process under complex reservoir conditions, and has important practical application value.