基于ESSA-LSTM的养殖工船水质溶解氧预测方法研究

Research on water quality dissolved oxygen prediction method based on ESSA-LSTM for aquaculture ships

为了准确预测水质参数中的溶氧量,采用长短时记忆网络(Long Short-Term Memory,LSTM)模型,提出一种增强型麻雀搜索算法(Enhance Sparrow Search Algorithm,ESSA)以改进预测率的精确性。该算法引入了Circle混沌映射进行种群初始化,并结合正弦余弦算法和Levy飞行策略分别对侦察者、跟踪者的位置进行更新,以促使麻雀个体能够快速跳出局部最优解。首先将ESSA与多种其他算法进行多形态基准函数对比测试,结果表明该算法在多个基准函数上展现出出色的性能和鲁棒性;随后将其应用于LSTM模型参数寻优,并与其他优化算法进行比较,结果显示基于ESSALSTM模型的预测率达到99.071%,相较于基本麻雀搜索算法(Sparrow Search Algorithm,SSA)、灰狼优化算法(Grey Wolf Optimizer,GWO)、海洋捕食算法(Marine Predators Algorithm,MPA)、鲸鱼算法(Whale Optimization Algorithm,WOA)分别提升了2.142%、6.653%、6.682%、7.714%。研究表明,使用ESSA显著提高了溶解氧预测率,并有效减少了参数设置的盲目性和时间成本。

In order to accurately predict the dissolved oxygen content in water quality parameters,we adopted a Long Short Term Memory(LSTM)model,and proposed an Enhanced Sparrow Search Algorithm(ESSA)to improve the accuracy of the prediction rate.Besides,to prompt individual sparrows to swiftly depart from the local optimal solution,the algorithm introduced Circle chaotic mapping for population initialization,and integrated sine-cosine algorithm and Levy flight strategy to update the positions of scouts and trackers,respectively.Firstly,we compared ESSA with various other algorithms for multi form benchmark function testing,and the results reveal that the algorithm exhibited excellent performance and robustness on mul-tiple benchmark functions.Subsequently,we used ESSA to explore LSTM model parameters and compared it with other optimi-zation strategies,and the results show that the prediction rate based on ESSA-LSTM model reached 99.071%,which was im-proved by 2.142%,6.653%,6.682%and 7.714%compared with basic Sparrow Search Algorithm(SSA),Gray Wolf Optimiza-tion Algorithm(GWO),Marine Predation Algorithm(MPA),and Whale Optimization Algorithm(WOA),respectively.The re-sults show that the use of ESSA significantly improves the prediction rate of dissolved oxygen(DO)and effectively reduces the blindness and time cost of parameter settings.