Page 147 - 《水产学报》2025年第8期

张铮，等 水产学报, 2025, 49(8): 089512

A PID prediction-based method for dissolved oxygen control in
industrial aquaculture

1
1
2
ZHANG Zheng , WU Changlin , ZHANG Zeyang , CAO Shouqi 1*
1. College of Engineering Science and Technology, Shanghai Ocean University, Shanghai 201306, China;
2. Shanghai City Electric Power Development Co., Ltd., Shanghai 200123, China

Abstract: Water quality is critical for aquaculture, with dissolved oxygen (DO) as a key parameter directly impacting organ-
ism survival, growth, and farming efficiency. In high-density intensive systems, frequent use of liquid oxygen for aeration to
maintain adequate DO increases costs. PID controllers are widely used for DO control due to their simple structure, high robust-
ness, and fast response. However, fixed parameters struggle with nonlinear systems, often causing performance degradation or
instability. Researchers have integrated intelligence into PID for online parameter tuning and combined predictive technology to
anticipate system dynamics, enhancing adaptability and reducing delays. This study first established a transfer function model
of aeration flow vs. DO concentration using experimental data, obtaining the DO control system model. It then proposed an
FSSCINET-QNN-PID controller—combining a fast-slow learning sample convolutional interactive network (FSSCINET) for
prediction with a quantum neural network (QNN) for PID parameter tuning-to improve response speed and anti-interference
capability. FSSCINET enhances DO prediction by integrating adapters (dynamic parameter adjustment) and memory modules
(capturing periodic changes) into SCINET, leveraging convolution and interaction for time-series data. QNN enables online
updates of PID parameters to handle nonlinear dynamics. A DO monitoring system in industrial aquaculture validated the
model. Results showed FSSCINET outperformed SCINET and FSNET, with MSE (0.037 5 mg/L), MAE (0.155 4 mg/L), and
RMSE (0.193 7 mg/L). FSSCINET-QNN-PID reduced adjustment time to 1 642 s with smaller overshoot compared to PID and
QNN-PID, stabilizing faster with minimal fluctuations. This study can provide a new idea for automatic regulation of DO in
factory farming.
Key words: industrial aquaculture; dissolved oxygen; control strategy; predictive PID
Corresponding author: CAO Shouqi. E-mail: sqcao@shou.edu.cn

Funding projects: Agricultural Science and Technology Innovation Project of Chongming District, Shanghai (2021CNKC-05-
06); Program for Shanghai Agricultural Science and Technology Innovation Project (I2023006); Shanghai Collaborative Innov-
ation Centre for Cultivating Elite Breeds and Green-culture of Aquaculture Animals (2021KJ02-12)

中国水产学会主办 sponsored by China Society of Fisheries https://www.china-fishery.cn
13