Term AI

TERM AI – 3D HEAT MAP OF A MARINE FARM USING AI

Julia Soares Menezes¹; Kennedy dos Santos Pacheco¹; Vitor Santos de Oliveira¹; Edson Anício Duarte¹(Orientador); Eduardo Galembeck² (Coorientador)

1 - Instituto Federal de São Paulo – Câmpus Campinas; 2 - Instituto de Biologia na Unicamp.

This project proposes the development of an autonomous system to remotely monitor water temperature in marine seaweed and shellfish farms, replacing manual measurements. Four meteo-oceanographic buoys will be used, each equipped with three DS18B20 sensors positioned at different depths, enabling the generation of a 3D map of thermal variation. Data, collected hourly, will be transmitted via LoRa to a central buoy, which will forward the information to the ThingSpeak platform using solar and hygroelectric energy. The objective is to ensure accurate, frequent, and low-cost measurements with long-distance coverage.

Temperature monitoring in marine farms helps estimate seaweed growth and preserve shellfish cultivation during periods of rising water temperature.

Manual monitoring requires physical effort, generates high costs, and presents limitations in precision and frequency. To address this issue, the Term AI project proposes a technological solution for remote and autonomous monitoring aimed at increasing efficiency, reducing costs, and ensuring production safety.

We are developing an embedded and autonomous system for temperature acquisition in marine farms that sends data to a cloud database and generates a three-dimensional heat map of the areas where marine farms are located. Artificial Intelligence resources will be used to predict variations and issue real-time alerts.

Aligned with United Nations Sustainable Development Goals 2, 8, 9, 12, 13 and 14.

MATERIALS AND METHODS

The system consists of four buoys, each equipped with three DS18B20 sensors, LoRa 32 microcontrollers, and a hybrid energy harvesting power system combining solar panels and hygroelectric cells (solar energy and hygroelectricity). Data are collected hourly and transmitted by the buoys to a central unit, where they are processed and sent to an online cloud storage platform called ThingSpeak.

Based on the stored database, the system software generates a 3D heat map for different depths, and through a proprietary Artificial Intelligence system, thermal variation predictions are made based on the analysis of these maps. The system was programmed in C++ using the Arduino IDE.

To develop the 3D heat map for temperature visualization, it was necessary to: collect data from the four buoys, recording position (x, y, and z, indicating depth and spatial location) and temperature from each of the three sensors; organize the data into NumPy arrays for processing; interpolate values using SciPy functions to estimate temperatures between measured points; and plot the results in 3D using libraries such as Matplotlib or Plotly.

For the AI component, a linear regression model was structured using TensorFlow, Pandas, and NumPy libraries. Both software implementations were developed in Python.

PRELIMINARY RESULTS

During project development, initial tests were conducted in water containers. After structuring a functional prototype, it will be implemented in a real environment to verify buoyancy, durability, and accuracy.

The three-dimensional heat map and AI-based analysis and prediction were tested using data obtained from bench tests and randomly generated numbers. Figure 6 refers to one of the bench tests, Figure 7 to the finalized web-based heat map, and Figure 8 to one of the AI training sessions using the imported data.

The system for remote temperature monitoring, including 3D heat map generation and AI analysis, has progressed promisingly, with plans for implementation in a marine seaweed and shellfish farm in partnership with the Fisheries Institute of Ubatuba. Once implemented, the project is expected to reduce workers’ effort and provide researchers with more detailed environmental data.

REFERENCES

CAVALLI, Ronaldo Olivera; FERREIRA, Jaime Fernando. The future of fishing and marine aquaculture in Brazil: mariculture. Ciênc. cult. (S. Paulo), v. 62, n. 3, p. 38–39, 2010.

WENXIAN, G.; WU, H.; XU, J.; DENG, W. PSO-BP neural network model for predicting water temperature in the middle of the Yangtze River. In: International Conference on Intelligent Computation Technology and Automation, vol. 2, pp. 951–954, IEEE, 2010.

ZOURMAND, A.; HING, A. L. K.; HUNG, C. W.; ABDULREHMAN, M. Internet of Things (IoT) using LoRa technology. In: IEEE International Conference on Automatic Control and Intelligent Systems (I2CACIS), 2019, p. 324–330. IEEE, 2019.

GERON, Aurélien. Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow: Concepts, Tools, and Techniques to Build Intelligent Systems. 2nd ed. Sebastopol: O’Reilly Media, 2019.