Wave height prediction method based on wavelet decomposition-neural network

Abstract

The invention discloses a wave height prediction method based on a wavelet decomposition-neural network and belongs to the field of marine monitoring technology. According to the method, a time sequence is generated through buoy data, the time sequence is decomposed and reconstructed, and clutter in the sequence is solved; neural network training sample processing is performed on the buoy data; aneural network model is trained, wavelet decomposition and reconstruction are performed on the time sequence, the number of training samples and wavelet decomposition layers is continuously optimizedduring training, and a feedback neural network is utilized to perform training to establish a corresponding wavelet decomposition-neural network model; and test samples established in the neural network are utilized to perform test sample testing on the wavelet decomposition-neural network model. Through the method, extreme wave height in a future period of time can be predicted to provide a powerful guarantee for offshore operation and offshore navigation.

Description

technical field [0001] The invention relates to the technical field of marine monitoring, in particular to a wave height prediction method based on wavelet decomposition-neural network. Background technique [0002] With the continuous development and utilization of land resources, the land resources are constantly depleted. Humans began to turn their attention to the resource-rich ocean, and the ocean has gradually become an important way for humans to obtain resources. However, the weather at sea is complex and changeable, and it is an important source of danger during marine operations. How to timely and accurately forecast extreme weather at sea, especially the extreme wave height of waves, has become the focus of research by experts and scholars today. [0003] The physical process of generating a series of waves by means of wind is very complex, unstable, non-linear, and uncertain. So far, the research on this physical process has not yet matured. The prediction of w...

AI Technical Summary

Problems solved by technology

Numerical methods are more accurate than parametric methods, can provide information at multiple locations simultaneously, and are more plausible when wind speed varies with its direction and area over a given duration, but require a large number of oceanographic and meteorological parameters , which increases the prediction cost in actual prediction and requires a large number of oceanographic and meteorological parameters, resulting in heavy data acquisition tasks and reduced operability

Images