Prediction of Cyanobacterial Bloom Based on Contrast Divergence-Long-Short-Term Memory Network

Abstract

The invention discloses a method based on contrast divergence. A method for predicting cyanobacteria bloom based on long-term and short-term memory network belongs to the technical field of water environment prediction. The method comprises steps of firstly, establishing a CD-LSTM model, using the improved moving average filtering algorithm to smooth the cyanobacterial bloom characterization factors; Then the influence factors of cyanobacteria bloom were reconstructed by using the contrast divergence algorithm. Finally, the LSTM model is used to predict the cyanobacteria bloom. The invention combines the contrast divergence algorithm, the moving average filter algorithm and the long-short memory network model, and provides a method based on CD-LSTM, which can improve the prediction accuracy of cyanobacteria bloom, provides a new idea for cyanobacteria bloom prediction.

Description

technical field [0001] The invention relates to a water bloom prediction method, which belongs to the technical field of water environment prediction. Specifically, it is a cyanobacterial bloom prediction method based on Contrastive Divergence-Long Short Term Memory networks (CD-LSTM). Background technique [0002] With the development of social economy, the phenomenon of eutrophication of water body is becoming more and more common, which has caused many troubles to people's life. Eutrophication of water body is a natural phenomenon that occurs in fresh water and is caused by the sudden excessive proliferation of algae due to the high content of nitrogen, phosphorus, and potassium in the water body. The algae bloom is a phenomenon in which a large number of algae grow or gather in a water body and reach a certain concentration, and it is the result of the combined effects of eutrophication and specific conditions in the water body. The occurrence of algae bloom seriously ...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to solve the problem that the existing water bloom prediction accuracy is not high, and the prediction of the highly nonlinear system cannot be processed only by a single factor or a single mathematical method. The present invention will compare the divergence algorithm, the moving average filter algorithm and the long-term Combined with the short-term memory network model, a CD-LSTM-based cyanobacterial bloom prediction method is provided, which can improve the prediction accuracy of cyanobacterial blooms and provide a new idea for cyanobacterial bloom prediction

Images