Abstract: Dissolved oxygen content is an important factor affecting eel culture. To improve the prediction accuracy of dissolved oxygen concentration in eel ponds, this paper proposes a model for predicting dissolved oxygen concentration in eel ponds based on the sparrow search algorithm (SSA) and long short-term memory neural network (LSTM), after optimizing the hyperparameters of LSTM model using SSA algorithm, the dissolved oxygen in circulating water eel culture ponds was predicted. The results showed that the prediction accuracy of SSA-LSTM model was 96.77%, which was 2.09%, 3.34% and 0.55% higher than the control models LSTM, GRU and PSO-LSTM, respectively. The other indicators of this model, , , and , were 0.67, 0.53, and 0.81, respectively, which also showed significant decreases compared with the control model. The results indicate that the SSA-LSTM model has good accuracy and robustness in predicting dissolved oxygen concentration in eel ponds, which can provide a basis for accurate regulation of water quality parameters in eel culture.

Key words: Dissolved oxygen prediction, Long and short-term memory neural networks, Sparrow search algorithm, Eel farming

摘要： 溶氧含量是影响黄鳝养殖的重要因素，为提高黄鳝池溶氧浓度的预测精度，提出一种基于麻雀搜索算法（SSA）和长短期记忆神经网络(LSTM)的黄鳝池溶氧浓度预测模型，即利用SSA算法优化LSTM模型的超参数后，对循环水黄鳝养殖池的溶氧浓度进行预测。结果显示：基于SSA-LSTM模型的预测准确率为96.77%，相较于对照模型LSTM、门控循环单元（GRU）、粒子群算法-长短期记忆神经网络（PSO-LSTM）分别提升了2.09%、3.34%、0.55%。该模型其他指标均方误差（）、平均绝对误差（）、均方根误差（）分别为0.67、0.53、0.81，相较于对照模型也有明显下降。研究表明，利用SSA-LSTM模型预测黄鳝池溶氧浓度具有良好的准确性和鲁棒性，可以为黄鳝养殖中水质参数精准调控提供依据。

关键词: 溶氧预测, LSTM, 麻雀搜索算法, 黄鳝养殖