Evaporation waveguide height short-term prediction method and system based on hyperparameter optimization

Abstract

The application relates to the technical field of evaporation waveguide height prediction, and discloses an evaporation waveguide height short-term prediction method and system based on hyperparameter optimization, which comprises the following steps: inverting historical meteorological observation elements to obtain a historical evaporation waveguide height time sequence; verifying the stationarity of the historical evaporation waveguide height time sequence, and pre-processing the historical evaporation waveguide height time sequence based on the verification result; constructing a framework for receiving time sequence data and outputting a prediction result, wherein the framework comprises a cycle processing logic with memory keeping and gate control functions; and performing automatic optimization on the hyperparameters of the framework based on a Bayesian inference strategy; through accurate inversion of historical meteorological elements of a target sea area and time sequence preprocessing, in combination with time sequence processing logic with memory keeping and gate control functions and a hyperparameter automatic optimization strategy based on Bayesian inference, the accuracy and stability of the evaporation waveguide height short-term prediction are significantly improved.

Images