A Low-Cost AI Buoy System for Monitoring Water Quality at Offshore Aquaculture Cages

Abstract

The ocean resources have been rapidly depleted in the recent decade, and the complementary role of aquaculture to food security has become more critical than ever before. Water quality is one of the key factors in determining the success of aquaculture and real-time water quality monitoring is an important process for aquaculture. This paper proposes a low-cost and easy-to-build artificial intelligence (AI) buoy system that autonomously measures the related water quality data and instantly forwards them via wireless channels to the shore server. Furthermore, the data provide aquaculture staff with real-time water quality information and also assists server-side AI programs in implementing machine learning techniques to further provide short-term water quality predictions. In particular, we aim to provide a low-cost design by combining simple electronic devices and server-side AI programs for the proposed buoy system to measure water velocity. As a result, the cost for the practical implementation is approximately USD 2015 only to facilitate the proposed AI buoy system to measure the real-time data of dissolved oxygen, salinity, water temperature, and velocity. In addition, the AI buoy system also offers short-term estimations of water temperature and velocity, with mean square errors of 0.021 °C and 0.92 cm/s, respectively. Furthermore, we replaced the use of expensive current meters with a flow sensor tube of only USD 100 to measure water velocity.

Keywords: artificial intelligence; machine learning; offshore aquaculture; water quality; wireless communications.

Figure 1

Architecture of the proposed offshore buoy.

Figure 2

Proposed offshore buoy: main body and flow sensor tube.

Figure 3

Operation flow of the proposed buoy.

Figure 4

Control box of the proposed AI buoy.

Figure 5

The multilaryer architecture and cell structure of long short-term memory (LSTM).

Figure 6

Prediction results of water temperature.

Figure 7

Prediction results of water velocity.

Figure 8

Regression of water velocity.

Figure 9

Smart water quality prediction, (a) current information, cage monitoring (at the top half part): water temperature, water velocity, dissolved oxygen, and salinity (from left to right and top to bottom), and weather information (at the lower half part): temperature, daily rainfall, humidity, wind speed, wind direction, maximum showers per hour (from left to right and top to bottom), and (b) prediction information, water temperature (at the top half part), and water velocity (at the lower half part).