Optimization of Profile Control and Oil Displacement Scheme Parameters Based on Deep Deterministic Policy Gradient

[Image: see text] The parameter design of profile control and oil displacement (PCOD) scheme plays an important role in improving waterflooding efficiency and increasing the oil field production and recovery. In this paper, the parameter optimization model and solution method of the PCOD scheme based on deep deterministic policy gradient (DDPG) are constructed with the half-year increased oil production (Q(i)) of injection well group as the objective function and the parameter range of PCOD system type, concentration, injection volume, and injection rate as constraints. Using the historical data of PCOD and extreme gradient boosting (XGBoost) method to construct a proxy model of PCOD process as the environment, the change rate of Q(i) of well groups before and after optimization is taken as the reward function; the system type, concentration, injection volume, and injection rate are taken as the action; and the Gaussian strategy with noise is taken as the action exploration strategy. Taking XX block of offshore oil field as an example, the parameters of the compound slug PCOD process (pre-slug + main slug + protection slug) of the injection well group are analyzed, that is, parameters such as the system type, concentration, injection volume, and injection rate of each slug system are optimized. The research shows that the parameter optimization model of the PCOD scheme established based on DDPG can obtain higher oil production PCOD scheme for well groups with different PCOD, and has strong optimization and generalization ability compared with the particle swarm optimization (PSO) model.