Method for predicting depth-averaged current velocities of underwater gliders based on data feature analysis

In this paper, the data feature of depth-averaged current velocities (DACVs) derived from underwater gliders is analyzed for first time. Two features DACVs have been proposed: one complex ingredients and small samples, other stationarity that occurs as length a DACV sequence increases. With these in mind, set methods combining statistical analysis machine learning are proposed to realize prediction DACVs. Four groups different sea trials South China Sea used verify method. Based on three general error criteria, performance model demonstrated. The persistence method comparison model. results show paper effective.