The one of the advantages about Laser Induced Breakdown Spectroscopy (LIBS) is multielement detection at the same time.In order to obtain the optimum signal in the multi-element measurements of water with LIBS, the present paper firstly models the numerical relationship between the signal-to-background ratio of characteristic spectral lines and the delay time and gate width time with BP neural network, using DM design experiment data as the checking sample to ensure the generalization ability of the BP neural network model. Based on the above model, genetic algorithm is used to optimize measurement parameters and the fitness function φ is defined. When the optimum delay time and gate width time is (15.5 μs, 21.5 μs), the minimum value of φ is 0.102 4. The optimization results of genetic algorithm are further confirmed with experimental results. So the method of parameters optimization overall improves S/B of multi-element measurements in water with LBS, and provides the reference for parameter optimization of other experiments.