Modelling and controlling dissolved oxygen in recirculating aquaculture systems based on mechanism analysis and an adaptive PID controller

doi:10.1016/j.compag.2021.106583

Computers and Electronics in Agriculture

Volume 192, January 2022, 106583

https://doi.org/10.1016/j.compag.2021.106583 Get rights and content

Highlights

•
This paper analyzed the influence of four main factors on the dynamic changes of DO.
•
The differential equation model of the DO control process was established.
•
The reliability of established model was verified in actual breeding environment.
•
An adaptive controller was proposed for DO control in recirculating aquaculture systems.
•
This work can provide theoretical basis and accumulate experience for practical application.

Abstract

In aquaculture, dissolved oxygen (DO) content is critical to the growth and development of aquatic products, and it must be precisely controlled. Based on the theory of microporous aeration mass transfer and mass conservation equations, this paper analyzed the influence of the four main factors of recirculating water flow, mechanical aeration, surface reaeration and respiration of shrimp on the dynamic changes of DO, and established a DO system dynamics model in the recirculating aquaculture system. Based on the established model, this paper proposed a fuzzy rule-optimized single neuron adaptive PID controller (FL-SN-PID) for precise control of DO. In order to verify the reliability of the established model, several aeration experiments with different aeration flows were conducted in Simulink and actual breeding system. All fitting R² between simulated data and measured data of the DO response are above 0.94, indicating that the established dynamics model can accurately approximate the actual breeding system and can provide a basis for the design of the controller. In the simulation experiment, the proposed controller was applied to the DO tracking control in three scenarios, and its tracking performance was also compared with the traditional PID controller and the SN-PID controller. In performance analysis, the integral of absolute error (IAE) and the integral of squared error (ISE) of the FL-SN-PID controller are obviously smaller than the other two controllers, indicating this controller has better control accuracy and robustness, and can be competent for precise control of DO in aquaculture.

Introduction

Recirculating aquaculture is the main production method for factory farming in the future, with great advantages such as high breeding efficiency and saving water resources (Yue & Shen, 2021). Accurate control of water quality has always been the top priority in recirculating aquaculture, among which DO is the most important water quality factor to measure the quality of water body(Saber et al., 2020, Cao et al., 2021). The DO concentration is within a reasonable range, which is suitable for optimal growth and development of aquatic products and improves breeding efficiency. Relevant studies have shown that too low DO content will severely restrict the healthy growth of aquatic products, and even cause widespread death in severe cases (Fijani et al., 2019). However, long-term high DO content can cause fish to suffer from air bubble disease, which is especially harmful to fish eggs and juvenile fish (Wei et al., 2019a, Wei et al., 2019b). At present, aquaculture plants basically adopt the production method of regular oxygen addition to regulate DO content, and it often leads to a phenomenon where the DO content is too low, which not only fails to meet the demand for precise regulation of DO, but also increases the power consumption. For this reason, precise control of DO with the help of modern intelligent technology is the key to reducing breeding risks and improving breeding benefits.

The precise control of DO is to control the output air flow of the aerator to ensure that the DO is always close to the setting value or kept within a given range (Yan et al., 2021). The key to precise control of DO is to establish a mathematical model of the controlled process and design a controller. The mathematical model of DO can accurately reflect the influence of various factors on the dynamic changes of DO. This model can be used to explore the intrinsic behavior of multiple factors that is close to the real process, and provide the necessary information and data for the design and debugging of the process control system in computer, which could greatly reduce the cost of design experiments and speed up the design process (Manap et al., 2021). Therefore, in order to achieve precise control of DO, it is first necessary to establish a mathematical model that can accurately approximate the actual DO control system. Based on this mathematical model, the controller reasonably calculates the control signal according to the DO deviation and acts on the aerator to make the DO content always close to the setting value.

In terms of mathematical modelling of the controlled process, it can be realized by two methods: experimental modelling method and mechanism modelling method. The experimental modelling method is generally only used to establish an input–output model, that is, a model obtained after some processing is performed on the measured data of the input and output of the experimental process. Its main feature is to treat the researched object as a black box and describe its dynamic characteristics completely from the external characteristics (Khan et al., 2018). Although the experimental modelling method has the advantages of simplicity and ease of operation, the model obtained by this modelling method cannot reflect the influence of the multiple factors on the dynamic changes of DO, such as the respiration consumption, atmospheric reaeration, circulating water flow, aeration operations, photosynthesis, etc. (Yin et al., 2021). The mechanism modelling method aims to study the interdependence between system behavior and internal mechanism, and establish a mathematical model to analyze the changing law of DO with changes in external factors (Reeder, 2011). Many researchers have used mechanism modelling methods to analyze the dynamics of DO in different systems (Mandal et al., 2012, Nagisetty et al., 2019). Since the mechanism model can fully reflect the quantitative influence of various physical or chemical factors on DO, the resulting models have the advantage of high accuracy. However, there is no research on the dynamic modelling of DO in the recirculating aquaculture system. In order to clarify the influence of various factors on the dynamic changes of DO, this paper adopts the mechanism modelling method to establish the mathematical model of the controlled system.

In terms of controller design, the proportional-integral-derivative (PID) controller is a commonly used controller, which has the advantages of simple structure, easy operation, and strong robustness (Moharam et al., 2016, Alcaina et al., 2019, Cao, 2020). However, the traditional PID control parameters cannot be adjusted online, so there are still some deficiencies in the control accuracy. To this end, many researchers combine neural networks with PID controllers, and use the powerful self-learning capabilities of neural networks to adjust PID parameters to improve the adaptive capabilities of PID, such as back propagation-PID (Huang et al., 2019, Zhang and Yuan, 2020) and radial basis function-PID controller (Qu et al., 2011, Attaran et al., 2016, Zhou et al., 2021). In addition, PID parameters adjusted based on fuzzy rules have also been proposed. This kind of controller is mainly based on expert experience to formulate a series of fuzzy rules to adjust PID parameters online to improve the adaptive ability of the controller (Wei et al., 2019a, Wei et al., 2019b). Compared with artificial neural network adjustment, the method based on fuzzy rule adjustment does not require a lot of complicated calculations, which has the advantage of higher efficiency. Although the controllers proposed above have achieved satisfactory control effects in their respective fields, most of them have the disadvantages of more adjustable parameters, difficult to tune and time-consuming, especially for neural network-based PID controllers. The aquaculture environment is complex and changeable, and the DO control system has nonlinearity and hysteresis, which determines that a controller with better performance needs to be used to achieve a good control effect in aquaculture. Traditional PID controller based on neural network, due to the complex internal network structure of neural network and many adjustable parameters, it is difficult to achieve satisfactory control effect in the actual aquatic environment. Therefore, to propose an adaptive controller with few adjustable parameters, simple network structure, and strong robustness has always been the goal pursued by engineering designer.

Based on this, this paper first analyzed the influence of circulating water flow, mechanical aeration, surface reaeration and respiration of shrimp on the dynamic changes of DO, and constructed a differential equation model of the dynamic changes of DO. Based on this model, a FL-SN-PID controller was designed. The controller combines a single neuron network with a PID controller to improve its adaptive ability, and then introduces fuzzy logic rules to realize the self-tuning of the gain coefficient K of the SN-PID controller. The contributions of this paper mainly include two aspects: (1) the method of mechanism modelling was used to establish a differential equation model that can accurately reflect the dynamic changes of DO, providing a basis for the analysis and design of the DO control system; (2) a FL-SN-PID controller was proposed, which improves the steady-state performance and control accuracy of the controller by introducing fuzzy logic rules to adjust the value of gain coefficient K.

The rest of this paper is arranged as follows: the second part gives the method of establishing the differential equation model of DO in the recirculating aquaculture system; the third part describes the design principle of the FL-SN-PID controller; the fourth part is the experimental results and discussion; the fifth part is the conclusion.

Section snippets

Experimental system description

Taking the aquaculture base of Ningbo Institute of Oceanography and Fisheries of Zhejiang Province as the experimental site, a small-scale recirculating aquaculture system was built, and its structure is shown in Fig. 1. The recirculating aquaculture system mainly consists of three parts, namely aquaculture tank, waste filtering device and ammonia nitrogen treatment device. The aquaculture tank can be regarded as a cylindrical container with a radius of 0.75 m and a water depth of 0.85 m. By

Single neuron adaptive PID controller

The single neuron adaptive controller has self-learning ability and can approximate any non-linear function. It realizes the function of self-adaptation and self-organization by adjusting the weighting coefficient (Ding et al., 2020). In order to improve the performance of conventional PID controllers in controlling complex and time-varying systems, this paper combines single neuron network and PID controller (ie SN-PID controller) to obtain more satisfactory control effects. The network weight

Model parameter determination from measured data

It can be seen from Section 2.3 of this article that the dynamic change of DO in the recirculating aquaculture system is mainly affected by four factors. Formula (10) is selected to describe the respiratory oxygen consumption rate of total shrimps, and the dynamic change of DO in aquaculture tank can be described by the following differential equation, as follows: $\frac{dC}{dt} = \frac{Q_{in}}{V} (C_{in} - C) + (k_{1} + k_{2} Q {(t - t_{d})}^{k_{3}}) β^{(T - 20)} (C_{sat} - C) - MR / V$

In the formula (18), the parameters in the section of circulating water flow can

Conclusion

In view of the time-varying and nonlinear characteristics of the DO control process in aquaculture, this paper utilized the mechanism modelling method to analyze the influence of multiple variables on the dynamic changes of DO, and established a DO system dynamic model in the recirculating aquaculture system. In the reliability verification of the model, the fitting R² between the simulated data and the experimental data under different aeration flow is all above 0.94, indicating that the

CRediT authorship contribution statement

Xinhui Zhou: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Writing – original draft, Writing – review & editing. Jianping Wang: Funding acquisition. Lin Huang: Data curation. Daoliang Li: Writing – review & editing. Qingling Duan: Conceptualization, Funding acquisition, Methodology, Project administration, Resources, Supervision, Writing – review & editing.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This work was supported by the Project of Shandong Province Major Scientific and Technological Innovation Project-Integration and Demonstration of Key Technology for Land and Sea Relay Fish Precision Farming (Grant No. 2019JZZY010703), Ningbo Public Welfare Science and Technology (Grant No. 202002N3034), and the Jiangsu Agricultural Science and Technology Innovation Fund (Grant No. CX(19)1003).

References (36)

J. Alcaina et al.
Delay-independent dual-rate PID controller for a packet-based networked control system
Inform. Sci.
(2019)
S.M. Attaran et al.
A novel optimization algorithm based on epsilon constraint-RBF neural network for tuning PID controller in decoupled HVAC system
Appl. Therm. Eng.
(2016)
X. Cao et al.
A three-dimensional prediction method of dissolved oxygen in pond culture based on Attention-GRU-GBRT
Comput. Electron. Agric.
(2021)
J.M. Chern et al.
Effects of impurities on oxygen transfer rates in diffused aeration systems
Water Res.
(2001)
Y.C. Ding et al.
Design and experiment of the single-neuron PID navigation controller for a combine harvester
Trans. Chin. Soc. Agric. Eng.
(2020)
E. Fijani et al.
Design and implementation of a hybrid model based on two-layer decomposition method coupled with extreme learning machines to support real-time environmental monitoring of water quality parameters
Sci. Total Environ.
(2019)
G. Huang et al.
A BP-PID controller-based multi-model control system for lateral stability of distributed drive electric vehicle
J. Franklin Inst.
(2019)
O. Khan et al.
Optimized PID controller for an industrial biological fermentation process
J. Process Contr.
(2018)
H.H. Manap et al.
Control for carbon dioxide exchange process in a membrane oxygenator using online self-tuning fuzzy-PID controller
Biomed. Signal Proces.
(2021)
S. Mandal et al.
Dynamic modelling of dissolved oxygen in the creeks of Sagar island, Hooghly-Matla estuarine system, West Bengal, India
Appl. Math. Model.
(2012)

A. Moharam et al.

Design of optimal PID controller using hybrid differential evolution and particle swarm optimization with an aging leader and challengers

Appl. Soft Comput.

(2016)

R. Nagisetty et al.

Dissolved oxygen modelling of effluent-dominated macrophyte-rich Silver Bow Creek

Ecol. Model.

(2019)

B.C. Reeder

Assessing constructed wetland functional success using diel changes in dissolved oxygen, pH, and temperature in submerged, emergent, and open-water habitats in the Beaver Creek Wetlands Complex, Kentucky (USA)

Ecol. Eng.

(2011)

E. Royer et al.

Estimating oxygen consumption of rainbow trout (Oncorhynchus mykiss) in a raceway: a Precision Fish Farming approach

Aquacult. Eng.

(2021)

R.F. Weiss

The solubility of nitrogen, oxygen and argon in water and seawater

Deep Sea Res.

(1970)

Y. Yan et al.

Fault prognosis of HVAC air handling unit and its components using hidden-semi Markov model and statistical process control

Energ. Build.

(2021)

H. Yang et al.

A quick condition adaptive soft sensor model with dual scale structure for dissolved oxygen simulation of recirculation aquaculture system

Comput. Electron. Agric.

(2019)

L. Yin et al.

Modeling dissolved oxygen in a crab pond

Ecol. Model.

(2021)

Cited by (20)

Economically optimal operation of recirculating aquaculture systems under uncertainty
2024, Computers and Electronics in Agriculture
This study presents a novel economical operation scheme for recirculating aquaculture systems (RASs). The proposed approach is comprised of a moving horizon estimator and an economic model predictive controller (EMPC), which, respectively, provide the necessary feedback for the EMPC and make economically optimal decisions for the RAS. The scheme is enabled by a mechanistic RAS model, which is coupled with an objective function that jointly optimizes the economics of aquatic biomass growth and utility expenditure. The proposed scheme is tested on a rainbow trout case study, whereby it is compared against the previous state-of-the-art control approach in aerator failure and temperature disturbance scenarios. In both test scenarios, EMPC scheme outperforms the previous state state-of-the-art controller in terms of process economics (up to $41 %$ improvement) as well as shorter batch times (up to $22 %$ shortening). Furthermore, the scheme is shown to be capable of estimating uncertain parameters with little error (generally in the order of $5 %$ ) and negligible effect on economic performance when compared to cases with full measurement access and no uncertainty. The proposed scheme provides a new and automated way to ensure RASs operate with maximal productivity, thus further incentivizing sustainability and the uptake of this technology.
Numerical investigations on dissolved oxygen field performance of octagonal culture tank based on computational fluid dynamics
2024, Computers and Electronics in Agriculture
In aquaculture, dissolved oxygen (DO) is essential for the growth and development of aquatic products, particularly the recirculating aquaculture system (RAS). Measurement and control of DO concentrations have been extensively researched, but the transport mechanisms have not been realized due to the lack of reliable flow transfer models. To solve this problem, computational fluid dynamics (CFD) coupled with Euler-Euler and Species Transport Model (CFD-Euler-Euler-STM) model was employed to investigate the spatial and temporal variation of DO in recirculating aquaculture tanks. In this study, a user-defined function (UDF) was used to couple the CFD-Euler-Euler-STM. In this study, the CFD-Euler-Euler-STM model can simulate the detailed 3D flow field distribution of DO (MAPE = 4.40 %) by comparing on-site measurement data with simulation results. The results show that the maximum DO concentration is 5.68 mg/L at a gas holdup of 0.098, and that higher relative velocity and flow velocity magnitude lead to higher dissolved oxygen concentrations. In addition, the DO concentration is consistent with the streamline distribution. Moreover, the model provides a promising tool platform for the accurate regulation of DO and aeration optimization in aquaculture.
Modelling biochemical oxygen demand in a large inland aquaculture zone of India: Implications and insights
2024, Science of the Total Environment
Water quality surveillance is tough, and a specific timely management is necessary for the inland aquaculture ponds and ecology as well. Real time quality monitoring involves the study of numerous parameters includes physical (turbidity, temperature, and specific conductivity), chemical (pH, calcium, manganese, chlorides, iron, biochemical oxygen demand), and biological (bacteria and algae). It is also crucial to recognize the inter-dependence among the parameters. Alternatively, these relationships can be predicted with statistical and numerical modelling. Organic strength parameter 5-day biochemical oxygen demand (BOD) is a significant parameter to evaluate since its impact is very high on the quality of water, aquatic life, and other biological concerns. This study focuses on the prediction of BOD using six traditional and four boosting algorithms considering ten input physicochemical attributes. The attributes were fine-tuned for highly precise predictions by removing extreme values from the data set using data outlier treatment. The prediction results are compared using performance metrics such as coefficient of determination (R²), root mean square error (RMSE), mean square error (MSE), and mean absolute error (MAE). The findings revealed that boosting algorithms outperform the results of traditional models with the highest prediction accuracy. Among the boosting algorithms, eXtreme Gradient Boosting algorithm (XGBM) is found highly appropriate for the inland aquaculture waters with R² = 0.95, RMSE = 0.31, MSE = 0.09, MAE = 0.1. Finally, this study provides a systematic evaluation of the BOD in the aquaculture waters and has a significant contribution to water management and eco-balance.
Deep learning with PID residual elimination network for time-series prediction of water quality in aquaculture industry
2023, Computers and Electronics in Agriculture
Time-series prediction of water quality is the most critical component of water quality monitoring in the aquaculture industry. Accurate multi-step ahead prediction of water quality can provide reasonable support for production decisions and reduce breeding risks. However, because of the nonlinearity and strong coupling between water quality factors, it is difficult to obtain accurate prediction results and reduce prediction errors. For solving this problem, we innovatively propose a methodology wherein deep neural networks (DNN) coupled with a proportional–integral–derivative residual elimination network (PID-RENet) are used for time-series prediction of water quality. In the proposed method, PID-RENet is mainly composed of two parts: a conventional PID controller and a backpropagation (BP) neural network. The role of the PID controller is to calculate the control amount according to the predicted historical deviation and to correct the prediction result of the benchmark DNN model. The BP neural network is used to dynamically adjust the three parameters of the PID controller for improving its adaptive ability. To verify the effectiveness and practicability of the proposed method, a series of multi-step ahead prediction experiments were performed on a publicly available dataset and two self-constructed water quality datasets. The results for the time-series prediction of all types of datasets suggested that the error performance indicators of the model corrected by PID-RENet were all better than those of the evaluated benchmark DNN models, indicating that PID-RENet can provide more accurate predictions.
Fault diagnosis method of dissolved oxygen sensor electrolyte loss based on impedance measurement
2023, Computers and Electronics in Agriculture
Dissolved oxygen affects the growth and survival of aquatic organisms and considers as an important indicator of aquaculture water quality. Moreover, the electrochemical dissolved oxygen sensor may become faulty due to flow shock or damage by organisms which destroys the dissolved oxygen permeable membrane and results in loss of electrolytes. In this paper, a method was developed for detecting electrolyte loss faults in polarographic dissolved oxygen sensors by directly measuring impedance of sensing electrodes without additional sensing elements or modifying the sensor. First, the factors interfering electrolyte concentration-impedance relationship and the requirement for temperature compensation was determined. Secondly, a fault diagnosis of internal electrolyte loss was achieved by using the difference between the theoretically calculated value and the impedance measurement value in combination with the electrode reaction law. Finally, the impedance detection method was demonstrated using artificially simulated faults. The results showed that the developed method had accurate tracking of electrolyte consumption with a Pearson coefficient of 0.987. Three different levels of breakage failures were simulated and detected as 4.48 %, 21.07 % and 73.47 %, respectively. The mean absolute error of the normal state was about 1.67, which increased to 16.3 after a period of failure. The developed method meets the requirements of embedded development because it uses a simple fault detection device and easy data processing using a simple microprocessor. Furthermore, the diagnosis result has been demonstrated as an indicator of membrane health and breakage conditions of dissolved oxygen sensors.
Closed-loop operation of a simulated recirculating aquaculture system with an integrated application of nonlinear model predictive control and moving horizon estimation
2023, Computers and Electronics in Agriculture
High water recirculation combined with a low rate of water exchange can deteriorate water quality and affect fish welfare in recirculating aquaculture systems (RAS). A nonlinear model predictive controller (NMPC) integrated with moving horizon estimation (MHE) was implemented in this work to maintain water quality parameters (e.g., oxygen, ammonia, and nitrate concentration) in RAS. The performance of the proposed control scheme was then evaluated in the presence of process uncertainty and measurement noise with full and partial accessibility to the system’s states under several common scenarios that can arise during the operation of RAS. In all scenarios, the proposed framework showed acceptable performance in keeping the water quality parameters close to their setpoints, demonstrating the high potential of this strategy in closed-loop operation of RAS.

View all citing articles on Scopus

View full text

Modelling and controlling dissolved oxygen in recirculating aquaculture systems based on mechanism analysis and an adaptive PID controller

Highlights

Abstract

Introduction

Section snippets

Experimental system description

Single neuron adaptive PID controller

Model parameter determination from measured data

Conclusion

CRediT authorship contribution statement

Declaration of Competing Interest

Acknowledgments

Inform. Sci.

Appl. Therm. Eng.

Comput. Electron. Agric.

Water Res.

Trans. Chin. Soc. Agric. Eng.

Sci. Total Environ.

J. Franklin Inst.

J. Process Contr.

Biomed. Signal Proces.

Appl. Math. Model.

Appl. Soft Comput.

Ecol. Model.

Ecol. Eng.

Aquacult. Eng.

Deep Sea Res.

Energ. Build.

Comput. Electron. Agric.

Ecol. Model.