Deep Learning and Streaming Data for Real-Time Water Quality Monitoring in Shrimp Aquaculture: A Systematic Literature Review

Abstrak

Shrimp aquaculture is a critical part of the global seafood industry but remains highly vulnerable to environmental fluctuations and disease outbreaks, largely due to inadequate water quality management. Recent advances in deep learning and real-time streaming data offer promising solutions for intelligent monitoring and control. This systematic literature review evaluates state-of-the-art approaches integrating deep learning models and streaming architectures for water quality monitoring in shrimp aquaculture. Following PRISMA guidelines, 38 peer-reviewed studies published between 2020 and 2025 were selected from Scopus, with emphasis on tropical aquaculture systems. The findings are organized into four themes: predictive deep learning models, smart sensing technologies and data sources, streaming and edge computing architectures, and behavioral monitoring with production outcomes. LSTM, GRU, and CNN models show strong performance in predicting dissolved oxygen, pH, and temperature. Data are commonly collected through IoT sensors, UAVs, and AI-based imaging systems, enabling high-speed acquisition and detailed spatial information. Streaming platforms such as Apache Kafka, combined with embedded AI and edge computing, support near-real-time analytics and responsive system operation. Visual behavioral monitoring further enables early detection of stress indicators and improves operational efficiency. Overall, integrating deep learning with streaming data can substantially enhance sustainability and efficiency in shrimp farming. However, challenges remain in standardization, scalability, and behavioral modeling. Future research should prioritize benchmarking, hybrid edge–cloud architectures, and long-term validation studies.