IoT and machine learning approach for the determination of optimal freshwater exchange rate in aquaponics

The rapid global population increase and rising food demand have driven interest in alternative agricultural methods that offer improved resource efficiency. Due to constraints in land, water, and energy availability, controlled environment production systems such as aquaponics are gaining attention as integrated approaches to producing fish and plants. While aquaponics is widely recognised for its efficient water reuse, recent research suggests that regulated water exchange may also provide operational benefits. Both water exchange and water recirculation, therefore, play an important role in maintaining stable conditions and supporting performance within aquaponic systems. Swiss chard is a commonly used leafy vegetable in aquaponics systems because it grows well in nutrient rich water, has a relatively short growth cycle, and is sensitive to water quality conditions. Its growth patterns reflect the overall health and stability of the aquaponics environment, making it an effective indicator for monitoring plant development. Therefore, in this study, Swiss chard plant data analysis was conducted to assess plant growth stages and predict harvest timing, demonstrating how optimised water management and environmental monitoring can directly influence plant productivity in aquaponics systems. While standard coupled aquaponics management relies on manual monitoring and empirical adjustments of water exchange, these approaches are often time consuming and may not capture the complex, dynamic interactions between water quality parameters and plant growth, as reported in previous studies. Literature also indicates that Internet of Things-enabled sensing and machine learning provide a data-driven approach to continuously monitor system conditions and predict optimal freshwater exchange, addressing the limitations of conventional methods and enabling more precise, efficient, and scalable aquaponics management. Various water quality and environmental sensors were deployed along with an Arduino UNO microcontroller to collect real-time data from the aquaponics system. Machine learning techniques, including regression, classification, and ensemble models, were applied for predicting optimal water exchange rates, classifying plant growth stages, and forecasting harvest timing. Specifically, models such as Decision Tree, Random Forest, Gradient Boosting, Ridge Regression, Polynomial Regression, XGBoost, GRU, Convolutional Neural Networks (CNN), and Seasonal Auto Regressive Integrated Moving Average (SARIMA) were employed to achieve accurate and robust predictions. This thesis addresses the central research question: "How can IoT and machine learning be applied to optimise freshwater exchange in aquaponics systems and support plant growth classification and harvest prediction to enhance system productivity?". To answer this question, four specific objectives were pursued. First, the optimal rate of freshwater exchange was determined using IoT-enabled monitoring and machine learning algorithms, evaluating the water quality index of fish water to stabilise conditions and support yield productivity. Second, a comparative analysis of standard aquaponics and vermiponic systems (incorporating earthworms) was performed, both with and without water exchange, revealing that regular water exchange increased productivity by 2.83% in standard systems and 5.54% in vermiponics (p<0.05). Third, plant growth stage classification was conducted using an aquaponic Swiss chard leaf image dataset to characterise development patterns within the system. Finally, machine learning models were employed to predict harvest timing based on plant growth stages and environmental data, providing farmers with actionable guidance to optimise harvest schedules according to market demand. Together, these objectives integrate water quality optimisation, system performance evaluation, and predictive plant-level analytics, demonstrating the benefits of consistent water exchange, the impact of earthworms, and the potential of IoT and machine learning to enhance aquaponics productivity. The study also proposes the design and implementation of a smart, cost-effective, modular, and expandable aquaponics system, providing significant guidance for aquaponics farmers. Overall, this study demonstrates that integrating IoT and machine learning for optimal freshwater management not only enhances water quality and system yield but also enables accurate monitoring of plant growth and harvest prediction, providing a comprehensive framework for improving productivity and decisionmaking in aquaponics systems under standard conditions.

Name dc:type.qualificationname

PhD

Grantor dc:publisher.institution

University of Wolverhampton

Year dc:date.issued

2026

Author dc:creator

• Chandramenon, Praveen

Advisors dc:contributor.advisor

• Tchuenbou-Magaia, Fideline
• Gascoyne, Andrew

• aquaponics
• vermiponics
• water exchange
• machine learning
• internet of things
• water quality
• smart technology

Licence

• https://wlv.openrepository.com/bitstreams/8d93db4b-4309-4367-a61b-1e50f71e86e4/download