DA-Bi-SRU for water quality prediction in smart mariculture

Due to the open nature of the mariculture environment, water quality factors are susceptible to the cross-influence of biology, physics, chemistry, hydrometeorology and human production activities. The changes of water quality parameters have the characteristics of non-linearity, dynamics, variability and complexity. We propose a novel water quality prediction model for pH, water temperature and dissolved oxygen, namely Double-Attention-Based Bidirectional Simple Recurrent Unit model (DA-Bi-SRU). First, we construct a new huge original dataset collected in time series, consisting of 23,000 sets of data. Then, the collected water quality parameters are sequentially preprocessed. Finally, we introduce a dual attention mechanism module for feature extraction and temporal sequences in the Bi-SRU model. Using the correlations between the water quality parameters and temporal dependencies information, the proposed model can significantly improve the accuracy of long-term prediction of water quality. The experimental results show that our DA-Bi-SRU model has higher prediction accuracy than the methods based on RNN (Recurrent Neural Network), LSTM (Long Short-Term Memory) and Bi-SRU, and the prediction accuracy can reach 93.06%. Therefore, in smart mariculture, farmers can know the changing trend of water quality in advance through our proposed method, and take timely countermeasures before the deterioration of aquaculture ecology. •The sampled water quality data is preprocessed. Then, we use the feature attention mechanism to calculate effects of other water quality parameters on the parameters to be predicted, thereby reducing the interference of irrelevant features on the model prediction.•The temporal attention mechanism is used to quantify the influence of the output information at different time points on the prediction results of the prediction time points. It can strengthen the expression ability of the output information at important times.•Based on the sparse deployment strategy, a wireless sensor network model for data acquisition and transmission in a wide area is constructed.