Aiming at the strong random volatility of ocean energy generation, power characteristic indexes, including amplitude stability, duration stability, and the coefficient of variation, were calculated for wave and tidal current power generation. A feature vector constructed from these power characteristic indexes was then used to perform dimensionality reduction and clustering of the time-series data of ocean energy generation power, thereby obtaining typical output scenarios of ocean energy generation.With the goal of optimizing the consumption rate, a model for optimizing the allocation of ocean energy generation capacity is constructed and transformed into a linear programming problem for solution.The calculation results based on actual data show that the optimized configuration results after clustering of characteristic indicators are basically consistent with the measured output data scenario, which verifies the effectiveness of the method.