Tidal turbine is a device which converts the kinetic energy of water into electric energy. The blade element momentum theory (BEM) is used to design the blade in this paper. Although tidal power resources are abundant, the actual operation of hydraulic turbines is not very good due to the limitations of turbine conversion efficiency and production cost. Therefore, this paper establishes a neural network model for variables and objective functions, and then applies multi-objective optimization algorithm to genetic optimization of the power coefficient, the main index of hydraulic performance of tidal turbines. The optimized results are verified by model test and sea trial. The results show that after optimizing the blade chord length distribution and pitch Angle distribution, the power coefficient of the turbine increases by 2%, and the optimal tip speed ratio range is also expanded, which is more conducive to the actual tidal turbine power generation, and has certain engineering significance.