基于CEEMDAN-ISSA-GRU混合的水质预测模型  

Hybrid Water Quality Prediction Model Based on CEEMDAN-ISSA-GRU

作　　者：马倩倩 赵丽琴 聂会 焦建格 MA Qian-qian;ZHAO Li-qin;NIE Hui;JIAO Gian-ge(College of Mechanical and Electrical Engineering,China Jiliang University,Hangzhou,Zhejiang 310018,China;Key Laboratory of Intelligent Manufacturing Quality Big Data Tracing and Analysis of Zhejiang Province,Hangzhou Zhejiang 310018,China;College of Water and Environmental Engineering,Zhejiang University of Water Resources and Electric Power,Hangzhou Zhejiang 310018,China)

机构地区：[1]中国计量大学机电工程学院,浙江杭州310018 [2]浙江省智能制造质量大数据溯源与应用重点实验室,浙江杭州310018 [3]浙江水利水电学院水利与环境工程学院,浙江杭州310018

出　　处：《计算机仿真》2025年第1期501-507,共7页Computer Simulation

基　　金：浙江省自然科学基金(LQ20E090006);浙江省自然科学基金(LZJWY22E090007);浙江省教育厅(Y201942482)。

摘　　要：准确预测河流水质可以有效解决水污染防治和水质监管等问题。然而,由于水质序列具有非平稳性、随机性和非线性,导致预测精度较低。现提出一种基于完全自适应噪声集合经验模态分解(CEEMDAN)、模糊熵(FE)和改进的门控循环单元(GRU)混合的水质预测模型。首先采用CEEMDAN将水质序列分解为若干个本征模态(IMF),并以FE为判据重构IMF序列,实现降噪目的。然后,利用改进的麻雀搜索算法(ISSA)确定GRU的超参数,提高GRU模型的性能和泛化能力。最后,将降噪后数据输入到ISSA-GRU模型进行预测。实验结果表明,与比较模型相比,所提出的模型具有更好的预测精度和误差性能,RMSE、MAPE、MAE分别为0.2518、0.1824和1.9441%,比基线GRU模型分别降低了40.93%、46.29%、46.41%。Accurately predicting river water quality can effectively solve problems related to water pollution prevention and quality monitoring.However,due to the non-stationarity,randomness,and nonlinearity of water quality sequences,the prediction accuracy is relatively low.This paper proposes a hybrid water quality prediction model based on a completely adaptive noise set empirical mode decomposition(CEEMDAN),fuzzy entropy(FE)and an improved gated recurrent unit(GRU).Firstly,CEEMDAN was used to decompose the water quality sequence into several intrinsic mode functions(IMF),and FE was used as a criterion to reconstruct the IMF sequence to achieve noise reduction purposes.Then,the improved sparrow search algorithm(ISSA)was used to determine the hyperparameters of GRU to improve the performance and generalization ability of the CRU model.Finally,the denoised data were input into the ISSA-GRU model for prediction.The experimental results show that the proposed model has better prediction accuracy and error performance compared to the comparative models.The RMSE,MAPE,and MAE are 0.2518,0.1824,and 1.9441%,respectively,which are reduced by 40.93%,46.29%,and 46.41%compared to the baseline GRU model.

关 键 词：水质预测 完全自适应噪声集合经验模态分解 模糊熵 门控循环单元 改进的麻雀搜索算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]