基于机器学习的养殖水体硝氮浓度预测方法研究进展

Advances in Machine Learning-based Methods for Nitrate Nitrogen Concentration Prediction in Aquaculture Water

摘要:

养殖水体硝氮（NO-3N）浓度是影响水产养殖生态平衡与经济效益的关键参数，其精准预测对水质调控至关重要。水体硝氮浓度变化受众多因素影响，精准预测极其困难。相比传统机理模型依赖复杂方程，机器学习方法能有效挖掘水质参数间的非线性关系，展现出显著优势。本文系统综述了机器学习驱动的养殖水体硝氮浓度预测技术与研究进展，包含传统机器学习方法、神经网络方法和混合机器学习模型，并阐述了各类方法性能与应用场景，在此基础上探讨了当前研究面临的挑战，并对未来研究趋势进行展望，为养殖水体硝氮浓度预警与调控提供方法参考。

Abstract:

The concentration of nitrate nitrogen (NO-3N) in aquaculture water bodies is a critical parameter influencing both the ecological balance and economic benefits of aquaculture. Accurate prediction of NO-3N levels plays a pivotal role in effective water quality control. The change of nitrate nitrogen concentration in water is affected by many factors, so it is difficult to accurately predict it. Compared with the traditional mechanism model which relies on complex equations, machine learning method can effectively mine the nonlinear relationship between water quality parameters, demonstrating significant advantages. The research progress of machine learning driven nitrate nitrogen prediction technology in aquaculture water was systematically summarized, and the performance and application scenarios of various methods were introduced. The traditional machine learning methods effectively captured the complex relationship between water quality parameters through feature engineering and integration strategy. The neural network method improved the prediction ability by time series modeling and attention mechanism, while the hybrid machine learning model broke through the limitations of a single model through algorithm collaboration and optimization strategy. Based on this, the challenges faced by the current research were discussed, and the future research trends were prospected, providing a method reference for the early warning and regulation of nitrate nitrogen concentration in aquaculture water.