均方误差
氧气
溶解有机碳
生化需氧量
环境科学
环境工程
计算机科学
水文学(农业)
生物系统
数学
统计
化学
环境化学
化学需氧量
工程类
岩土工程
生物
废水
有机化学
作者
Xinkai Cao,Ni Ren,Ganglu Tian,Yuxing Fan,Qingling Duan
标识
DOI:10.1016/j.compag.2020.105955
摘要
Pond culture is an open water body, the distribution of dissolved oxygen in water is three-dimensional. The demand for dissolved oxygen in aquatic products living in different water layers is different. The traditional one-dimensional prediction at one single monitoring point can‘t reflect the real situation of dissolved oxygen in different spaces in the pond. To solve these problems, a three-dimensional prediction method of dissolved oxygen based on Attention-Gated Recurrent Unit (GRU) - Gradient Boost Regression Tree (GBRT) was proposed in this paper. Firstly, the environmental factors affecting the distribution of dissolved oxygen were collected, and the dissolved oxygen prediction model of the central monitoring point was constructed using Attention-GRU. The three-dimensional coordinate system with the central monitoring point as the origin was then established, and the GBRT algorithm optimized by the Random Search algorithm(RS) was used to predict the dissolved oxygen in any position of the pond water. In the one-dimensional prediction of dissolved oxygen at the central monitoring point, the Attention-GRU model proposed in this paper had MSE of 0.121, MAE of 0.219, and RMSE of 0.348, which was a big improvement compared with LSTM model, ELM model and CNN model. In the three-dimensional prediction of dissolved oxygen in the pond, the RS-GBRT model proposed had MSE of 0.097, MAE of 0.191, and RMSE of 0.313. Compared with the models such as ExtraTree model, RandomForest model, and Bagging model, each evaluation index had been greatly improved. The experimental results indicated that the proposed method can accurately predict the dissolved oxygen in the three-dimensional space of the pond.
科研通智能强力驱动
Strongly Powered by AbleSci AI