Cyanobacterial bloom prediction method based on dynamic deep belief network

Abstract

The invention discloses a cyanobacteria bloom prediction method based on a dynamic deep belief network, belonging to the technical field of water quality monitoring. The prediction method of the present invention includes three parts: establishing a DDBN model, updating a learning algorithm of a dynamic restricted Boltzmann machine DRBM, and adopting a BP neural network reverse propagation algorithm to fine-tune the parameters of the DDBN model. The present invention improves the traditional RBM weight update formula, so that the weight, bias and learning rate update formulas are sequential, which facilitates the processing of time series problems and improves the prediction accuracy; the backpropagation algorithm only fine-tunes the parameters related to time t , which not only avoids the occurrence of local optimum, but also reduces the time for fine-tuning.

Description

technical field [0001] The invention relates to a water bloom prediction method, which belongs to the technical field of water quality monitoring. Specifically, it is a water bloom prediction method that improves the prediction accuracy of a dynamic deep belief network (Dynamic Deep Belief Nets, DDBN) prediction model established after analyzing the generation process of water blooms. Background technique [0002] In recent years, due to people's large-scale production, water eutrophication has become more and more common, causing serious water ecological problems. Water bloom is one of the typical characteristics of water eutrophication. The outbreak of water bloom destroys the structure of ecosystem and seriously restricts economic construction and social development. Water bloom has become a major problem in domestic and foreign governance. Therefore, it is of great significance to deeply study the outbreak process of algal blooms, and to effectively predict and simulat...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to solve the problems of the existing water bloom prediction accuracy is not high, it is difficult to predict through appropriate samples, and combine the improved depth belief network method with the time series model to build a dynamic depth belief network prediction model , so as to improve the prediction accuracy of algal blooms and provide a new idea for the prediction of algae blooms in lakes and reservoirs

Images