The invention discloses an initial rainwater storage pond optimization design method. On the premise of guaranteeing the water quality of the urban water system, taking the minimum total cost of waterquality guarantee of the water system as an optimization target, taking the total volume of all the storage tanks as a constraint condition, the optimal sewage interception rate of each storage tankis obtained by utilizing a mathematical model between the sewage interception rate of each storage tank and the total cost trained by an artificial neural network model and adopting a particle swarm optimization algorithm, and the design volume of each storage tank is obtained in combination with the relationship between the sewage interception rate of each storage tank and the runoff cumulant. The volume of the storage pond can be scientifically, economically, quickly and efficiently determined, the total cost is saved under the condition that limited conditions are met, and unnecessary wasteof manpower, material resources and financial resources is avoided.