A double-objective prediction and optimization method for buoys performance based on the artificial neural network

Determining the optimal geometry of the buoy is an important step in the design of oscillating wave energy converters, its main problem of which is the complexity and computational cost of the buoy shape parameterization process. To solve the above problems, the Genetic Algorism Back Propagation (GABP) neural network is used to optimize the longitudinal section of the buoy, and the Bessel equation is used to simplify the coordinates. 280 buoys with regular shapes are used as the training set, and 5% of them are selected as the test set. The trained neural network can accurately predict both the maximum heave displacement and additional mass of the buoy under a uniform wave condition, with a prediction accuracy of 93%. An optimal buoy geometry shape is constructed by randomly varying Bessel control points which enables the trained neural network to perform doubleobjective optimization of the buoy longitudinal section. The method proposed in this study can provide a new idea for the optimization of buoy geometry shape and performance prediction.