In Northern Europe, land-based aquaculture has emerged as a crucial component in securing sustainable animal protein supplies for the human food chain. Among various farming methods, Recirculating Aquaculture Systems (RAS) have gained significant attention due to their environmentally friendly and sustainable characteristics. However, the industry faces persistent challenges in production efficiency and fish health management that hinder its full potential.

Currently, most RAS operations rely heavily on manual observation, where farm workers visually assess fish feeding activity and behavior to determine production status. This experience-based approach demands substantial human resources, time, and requires operators to possess exceptional observational skills and extensive experience to accurately interpret fish health signals.

While modern sensor technology can collect vast amounts of data regarding water quality, feed consumption, and health parameters, this valuable information often remains siloed across different systems, failing to realize its full potential for real-time monitoring of fish health, welfare, and growth conditions.

An innovative project aims to fundamentally transform RAS management through artificial intelligence and statistical modeling. The initiative focuses on addressing core challenges in RAS production: feed management, feeding efficiency, and waste reduction. While video monitoring systems have become commonplace in marine aquaculture for observing fish behavior, their application in RAS systems remains largely unexplored.

This project seeks to bridge that gap by employing deep learning technology to analyze extensive video sequences of fish populations. The system is trained to "learn" and recognize fish behavior patterns, enabling intelligent feeding and health monitoring capabilities.

Feed constitutes a significant portion of aquaculture costs and plays a critical role in fish growth and water quality. Traditional feeding methods often lead to overfeeding, resulting in waste, increased operational costs, and potential water quality deterioration.

The project is developing a closed-loop intelligent feeding system that monitors fish feeding behavior in real-time. Using AI algorithms, the system analyzes hunger levels, feeding efficiency, and responses to adjust feeding amounts and frequency dynamically. This approach ensures fish receive optimal nutrition when needed, maximizing feed utilization while significantly reducing waste.

For instance, by analyzing feeding videos, the AI can determine whether fish are actively consuming feed. If activity appears low, the system reduces portions to prevent uneaten feed from accumulating. Conversely, when fish demonstrate strong feeding responses, the system can increase portions to support growth and nutrition.

Fish health represents the lifeline of RAS operations, where disease outbreaks can have devastating consequences. The project places particular emphasis on early health monitoring and warning systems. Through deep learning analysis of fish behavior, the AI can detect subtle abnormalities that might escape human observation.

Changes in swimming patterns, unusual group formations, or alterations in feeding behavior may signal emerging health issues. The AI system continuously analyzes these behavioral indicators, providing immediate alerts when potential risks are detected. This enables prompt intervention before problems escalate, significantly improving stock health and welfare.

Modern RAS facilities generate vast amounts of multidimensional data, including water quality parameters (dissolved oxygen, pH, ammonia, nitrites), feed consumption, growth metrics, and behavioral data from sensors and video monitoring. However, this information often remains fragmented across different management systems.

The project implements advanced data fusion techniques to integrate information from various sources. AI models process these diverse datasets to create comprehensive models of fish health and growth. This data-driven decision support system provides unprecedented insights for more scientific management decisions.

By correlating water quality data with behavioral observations, the system can identify relationships between environmental factors and fish health. Similarly, combining growth data with feeding information allows evaluation of different feeding strategies on growth rates and feed conversion ratios.

The technologies developed through this project demonstrate broad applicability and scalability for RAS facilities worldwide. By improving fish health, welfare, and production efficiency, these innovations contribute significantly to the sustainable development of aquaculture. The integration of AI promises to transform current practices, driving the industry toward smarter, more efficient, and environmentally responsible operations that support global food security.