Aquatic product culture water quality prediction method based on deep learning

A technology for water quality prediction and aquaculture, applied in neural learning methods, testing water, biological neural network models, etc., can solve problems that cannot satisfy complex nonlinearities

CN106198909AActive Publication Date: 2016-12-07CENT SOUTH UNIV
5 Cites 39 Cited by

Patent Information

Authority / Receiving Office
CN · China
Current Assignee / Owner
Publication Date
2016-12-07

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

The invention discloses an aquatic product culture water quality prediction method based on deep learning. By building a deep learning network having three-layer limited Boltzmann machine (RBM) and a layer of BP neural network, water quality sample data is used for training three limited Boltzmann machines by specific dispersion learning for extracting the deep characteristics of the water quality sample data, deep learning network parameter is optimized through BP, so that training on the deep learning network is complete. The trained deep learning network is used to the current water quality sample data, and the water quality prediction can be obtained on an output layer. The method can obtain the characteristic relevance between different water quality factors, and the water quality prediction accuracy is increased.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] The invention relates to a method for predicting aquaculture water quality based on deep learning. Background technique

[0002] my country's aquaculture industry is gradually developing from the traditional extensive stocking model to the industrialized and intensive farming model. Due to the high density of intensive aquaculture, when water quality problems occur, irreparable losses have often been caused, so the prediction of water quality has become the most critical part of intensive aquaculture.

[0003] The prediction of water quality in aquaculture refers to the acquisition of relevant information in the aquaculture process based on sensors or manual input, such as the current dissolved oxygen, pH, conductivity, ammonia nitrogen, temperature, water flow, water depth, pool area, number of fish, fish It can calculate the quality of water or the value of various parameters after a fixed period of time according to the growth stage of the fish, th...

Examples

Embodiment Construction

[0036] see figure 1 , figure 2 , the following embodiments illustrate the implementation process of the present invention from the two steps of training and prediction of specific examples.

[0037] (1) Training steps:

[0038] Step 1: Obtain water quality factors in aquaculture ponds through wireless sensor network nodes, including dissolved oxygen, pH, ammonia nitrogen, and temperature. Through the wireless transmission protocol, the data is transmitted to the server. The water quality evaluation data is obtained by using the water quality detector as the water quality marking data. At the same time, record the water depth, pond area, number of fish, and fish species of the pond on the server side. Use the above recorded data as training samples to train and study the deep learning network.

[0039] Step 2: Initialize the deep learning network. The input layer of the network (the visible layer of the first layer RBM) has 8 nodes, corresponding to the factor input in s...