Sistema IoT para monitoreo y control automático del agua en piscinas camaroneras con aprendizaje predictivo

Abstract

CONDITIONING OF PROJECT PUBLICATION. The high mortality rate in shrimp farming (Litopenaeus vannamei), primarily caused by unforeseen drops in dissolved oxygen levels, poses a critical challenge for the aquaculture industry. This project addresses this issue through the development of ShrimpGuard, a predictive monitoring and control system. The solution integrates Internet of Things (IoT) technology for real-time acquisition of temperature and oxygen data, utilizing the MQTT protocol for wireless communication and a SQLite database for information persistence. The core of the system is a machine learning model, based on the Random Forest algorithm, trained to predict oxygen levels with a four-hour horizon. A RESTful API, developed with FastAPI, manages access to data and predictions, which are visualized on an interactive web platform built with Streamlit. This platform not only displays metrics and graphs but also incorporates a user management system with roles and permissions. The results demonstrate that the system effectively anticipates hypoxia events, enabling timely interventions that significantly reduce shrimp mortality and optimize cultivation sustainability. Keywords: Aquaculture, IoT, Machine Learning, Random Forest, Predictive Monitoring, Water Quality, Litopenaeus vannamei.