A two-stage method is developed to solve a new class of multi-storage tank multi-source (MTMS) systems. In the first stage, the optimal storage policy of each tank is determined according to the electricity tariff, and the ground-level storage tank is modeled as a node. In the second stage, the genetic algorithm, combined with a repairing scheme, is applied to solve the pump scheduling problem. The objective of the pump scheduling problem is to ensure that the required volume is adequately provided by the pumps while minimizing the operation cost (energy cost and treatment cost). The decision variables are the settings of the pumps and speed ratio of variable-speed pumps at time steps of the total operational time horizon. A mixed coding methodology is developed according to the characteristics of the decision variables. Daily operation cost savings of approximately 11% are obtained by application of the proposed method to a pressure zone of S. Y. water distribution system (WDS), China. A two-stage method is developed to solve a new class of multi-storage tank multi-source (MTMS) systems. In the first stage, the optimal storage policy of each tank is determined according to the electricity tariff, and the ground-level In the second stage, the genetic algorithm, combined with a repairing scheme, is applied to solve the pump scheduling problem. the pumps while minimizing the operation cost (energy cost and treatment cost). The decision variables are the settings of the pumps and speed ratio of variable-speed pumps at time steps of the total operational time horizon. A mixed coding methodology is developed according to the characteristics of the decision variables. Daily operation cost savings of approximately 11% are obtained by application of the proposed method to a pressure zone of SY water distribution system (WDS), China .