Abstract: The traditional BP neural network algorithm is applied to the acquisition power prediction of wave power generation system,which is easy to fall into local optimum and insufficient generalization ability.Therefore,an improved particle swarm optimization neural network algorithm is proposed to dynamically adjust the learning factor and add mutation operators.Using indirect prediction strategy to build a direct-drive wave power generation system model from wave data to wave capture power,applying an improved algorithm to predict and analyze the wave history data,input the construction model,and obtain the wave capture power prediction value.Comparing and analyzing the simulation results of different prediction steps and different algorithms,it is found that the improved algorithm can effectively overcome the shortcomings of traditional algorithms and improve the prediction accuracy.