Abstract: Reservoirs optimal operation can improve the efficiency of water allocation and supply.Gongming water supply project is taken as an example, which is subsystem of Shenzhen urban water supply.The project constituted by hybrid reservoirs,including Gongming reservoir, Qiankeng reservoir, Ejing reservoir and Shiyan reservoir.In this project, Gongming reservoir is used to prevent continuous dry years or serious water pollution incidents, which might take great threat to urban water supply.Thus, the storage water of Gongming reservoir is taken as reserved water source for Shenzhen city water supply.For this special task, the Gongming reservoir need keeping at a relatively high water storage level for a long time, and the reservoir requires a certain amount of exchange water to maintain water quality health.The water supply benefit of the hybrid reservoirs and the amount of exchange water of Gongming reservoir are the key objectives.To study the optimal operation of Gongming water supply project for inter-basin water diversion and water supply operating, firstly, aggregation method is applied to aggregate the reservoirs into a “virtual reservoir”, which is used for simplifying topological structure of hybrid reservoirs.The dead storage and usable storage of “virtual reservoir” is formed by superposing the dead storage and usable storage of hybrid reservoirs respectively.Then based on “virtual reservoir”, the hybrid reservoirs operation rules for water diversion and water supply are established.Further, the multi-objectives genetic algorithm NSGA-II is applied to optimize the operation rules, and the multi-objectives of the operation are the minimum amount of diversion water and the maximum amount of supply water.The water supply operation rule curves divide the storage of “virtual reservoir” into 3 functional areas, which are normal water supply area, reduced water supply area and reserved water supply area.When total water storage of “virtual reservoir” keeps during normal water supply area, the urban demand water can be supplied adequately.When total water storage of “virtual reservoir” keeps during reduced water supply area, it indicates that the reservoirs are lacking of water, urban demand water can not be satisfied and water supply needs to reduce appropriately.The third situation is when continuous dry years or serious water pollution incidents occur, reserved water of Gongming reservoir is used to satisfy the urban demand water.The water diversion operation rule curves divide the storage into 2 functional areas, which are reduced water diversion area and normal water supply area.When total water storage of “virtual reservoir” keeps during reduced water diversion area, it indicates that the reservoirs have enough water to satisfy urban demand.The reserves reduce diversion water to prevent abandoned water.However, when total water storage of “virtual reservoir” keeps during normal water supply area, the reserves are lacking of water storage that should do water diversion.Pareto optimization technique is embedded in NSGA-II, which makes NSGA-II deal with multi-objectives at the same time.Firstly, Pareto optimization technique is applied to obtain 400 multi objective feasible solution sets that can meet water supply guarantee rate.In these sets, 183 optimal feasible solutions with different weights are selected.Then according to the 183 Pareto solutions, operation processes of water diversion and water supply are simulated.Based on the simulated results, the amounts of water diversion, exchange water, abandoned water and guarantee rate are counted as indicators, and the fuzzy method is used to analyze the relative membership degree of 183 Pareto solutions.In the scheme selecting process, water diversion and exchange water are assumed to be equal.Thus the highest relative membership degree of solution B is selected as satisfying scheme for real-time operation.Comparing and analysis the results, it demonstratesthat optimization operation can improve efficiency of water supply and water exchange.