Abstract: Due to the strong heterogeneity between layers and the prominent multi-objective conflict in complex reservoirs, it is difficult to achieve balanced injection and production and efficient development with the traditional division scheme of layers. Firstly, taking an actual reservoir as the research object, a multi-constraint multi-objective reservoir segment partition model was constructed with the objective of minimizing the heterogeneity index within the layer under the strict constraints of multi-dimensional process criteria. Secondly, aiming at the problem that the feasible region of the layer division model is extremely narrow and the traditional algorithm has low efficiency of static resource allocation, an evolutionary game-driven constrained multi-objective optimization algorithm is proposed. By introducing feasible maturity and replication dynamic equation, the algorithm realized the revenue driving and adaptive resource allocation among exploration, exploitation and diversity maintenance strategies, which significantly enhanced the optimization ability, convergence performance and diversity maintenance ability of the algorithm under complex constraints. Finally, the proposed algorithm was used to solve the model. The experimental results show that the optimization scheme generated by the proposed algorithm reduces the coefficient of variation of permeability by about 54%, the difference of moisture content by more than 85%, and the coefficient of variation of seepage resistance by about 52%. The optimization scheme significantly reduces the heterogeneity index in the layer, and improves the relationship between the coordination and layers and the injection and production. It provides a good technical support for the fine development of complex reservoir and the efficient exploitation of remaining oil potential.