LSTM-based sea surface temperature prediction method

Abstract

The invention discloses an LSTM-based seam surface temperature prediction method. The method comprises the following steps of: generating a prediction model and forecasting a sea surface temperature in a future time frame, wherein the step of generating the prediction model comprises the following sub-steps of: 1) carrying out z-score standardization on history sea surface temperature data in a set longitude and latitude range and generating a corresponding time column, 2) training an LSTM model by utilizing the time column and the z-score standardized history sea surface temperature data so as to obtain an LSTM-based sea surface temperature prediction model, wherein a time step length is 1 (day) and an input dimensionality is 1; and the step of forecasting the sea surface temperature in the future time frame comprises the following sub-steps of: inputting the z-score standardized intraday sea surface temperature data into a forecasting model, setting a prediction step number to obtainan output result of the future time frame, and carrying out reverse z-score standardization on the output result to obtain a sea surface temperature predicted value, wherein an output dimensionalityis 1. The method is capable of mining long and short time dependency relationships between data, is more suitable for learning long periodic change laws of sea surface temperatures, and is capable ofobtaining relatively good prediction results.

Description

technical field [0001] The invention relates to the field of marine meteorological forecasting, in particular to a sea surface temperature forecasting method based on an LSTM recursive neural network model. Background technique [0002] Sea surface temperature is a kind of marine meteorological data that is closely related to human beings. Production operations such as ship navigation, marine fishery production, and offshore oil platforms are closely related to sea surface temperature. It has certain guiding significance. The acquisition of sea surface temperature is far away from the inland, and traditional observation methods such as buoys are difficult to fully cover. At present, remote sensing satellites are mainly used for inversion, so the prediction of sea surface temperature has become relatively difficult. [0003] LSTM (Long-Short Term Memory) recurrent neural network has made major breakthroughs in many fields in recent years. This neural network can learn the lo...

AI Technical Summary

Problems solved by technology

[0004] The invention provides a sea surface temperature prediction method based on LSTM, which

Images