Previous Articles     Next Articles

Prediction model of aquaculture water temperature and pH based on BP neural network optimized by particle swarm algorithm

  

  1. (1 National Engineering Research Center for Information Technology in Agriculture ,
    Key Laboratory of Agri-informatics, Ministry of Agriculture,
     Beijing Engineering Research Center of Agricultural Internet of Things, Beijing 1000000,China
    2 College of Information and Technology, Taiyuan University of Technology, Taiyuan 030000, China;)
  • Online:2022-09-01 Published:2016-02-20

基于粒子群优化BP神经网络的水产养殖水温及pH预测模型

  

  1. （1国家农业信息化工程技术研究中心，农业部农业信息技术重点实验室，北京市农业物联网工程技术研究中心，北京 100000
    2太原理工大学信息工程学院，山西 太原 030000；）
  • 通讯作者: 孙传恒(1978—)，男，研究员，博士，研究方向：农业信息化。E-mail：sunch@nercita.org.cn.
  • 作者简介:徐大明(1991—)，男，硕士研究生，研究方向：算法优化，智能控制。E-mail：xudaming@vip.qq.com；
  • 基金资助:
    国家863计划项目（2012AA101905-02）；北京市自然科学基金资助项目（6152009）；国家现代农业产业技术体系建设专项 (CARS-49-03A）

PDF (PC)

Abstract: Focused on the problem of inaccurate aquaculture water temperature and pH prediction, a mixed algorithm for water quality parameters prediction which was based on particle swarm optimization BP neural network (PSO-BPNN) was proposed. Firstly, the particle swarm optimization (PSO) algorithm was applied in calculating the initial weights and thresholds of BP neural network (BPNN).Secondly, the abnormal data were fixed andthe six parameters of water quality as inputs were used, the temperature and pH value of the next time point were used as outputs to establish aquaculture water quality prediction model. Finally, the collected water quality data were used to conduct training in BP neural network, and the feasibility and performance of water quality prediction model was tested through experiments. Compared with support vector regression (SVR) and normal BP neural network, in the aspect of predicting water temperature using PSO-BPNN, the decreasing amplitudes of RMSE were 64% and 80% respectively, while in the aspect of predicting pH value, the decreasing amplitudes of RMSE were 32% and 65% respectively. The results of experiments show that aquaculture water quality prediction model based on PSO-BPNN is flexible, simple, convenient and it also has a good capacity of prediction.

Key words:  particle swarm optimization (PSO), BP neural network (BPNN), aquaculture, fishing condition warning, water quality prediction model

摘要： 针对养殖水质、水温及pH预测准确性低的问题，提出了一种基于粒子群优化BP神经网络的养殖水质参数预测方法。首先应用粒子群算法优化得出BP神经网络的初始权值和阈值，然后对得到的数据进行预处理，修复异常数据信息，再以当前时间的多个水质参数作为输入，下个时间点的水温、pH作为输出，建立养殖水质预测模型，最后利用采集的水质数据在BP神经网络中进行训练，并通过实验检验水质预测模型的可行性和预测性能。与支持向量回归（SVR）和传统BP神经网络相比，基于粒子群优化的BP神经网络在预测水温方面，均方根误差（RMSE）下降幅度分别为64.4%和86.7%；而在预测pH方面，RMSE下降幅度分别为11.1%和78.9%。研究表明，基于粒子群优化的BP神经网络养殖水质预测模型具有灵活简便、预测精度高、易于实现的特点，同时具有很好的预测能力。

关键词: 粒子群算法, BP神经网络, 水产养殖, 渔情预警, 水质预测模型