Abstract: Multi-objective optimal reservoir operation modeling is an effective method to deal with the problem faced by reservoir decision makers ? how to coordinate the benefits among power generation, irrigation and eco-environment of the reservoir. This study develops a multi-objective model for reservoirs that are operated with conflicting objectives of maximizing energy production and the water demand satisfaction rates for irrigation and ecological flow, taking into account the reliability of power generation. To solve the model, we present a new multi-objective optimization method of variable penalty coefficient (VPC). And a well-distributed Pareto frontier of multi-objective non-inferior solutions is obtained by screening from the feasible solutions. Comparison with the non-dominated sorting genetic algorithm Ⅱ (NSGA-Ⅱ) shows that this VPC algorithm performs better in seeking non-inferior solutions. Application in a case study shows that the algorithm and multi-objective model are satisfactory in adaptability and applicability to reservoir operation decision.