【作者】 韩旭； 段中兴； 尚旭东；

【Author】 HAN Xu;DUAN Zhongxing;SHANG Xudong;College of Information and Control Engineering Xi’an University of Architecture and Technology;

【机构】 西安建筑科技大学信息与控制工程学院；

【摘要】 针对水质指标波动大、非线性的特征导致水质预测精度低的问题,本文提出一种基于多头注意力机制的卷积神经网络-双向门控循环单元(CNN-BiGRU)水质预测混合模型。首先,对各项水质指标进行相关程度计算与筛选,筛选显著影响水质参数的指标作为输入变量。其次,融合CNN在特征提取和BiGRU在时间序列预测上的优势,搭建CNN-BiGRU混合模型,并引入了多头注意力机制(MHA)。最后,在CNN-BiGRU-MHA模型中输入特征矩阵实现水质预测。以国家公开水质数据作为实际算例,以氨氮预测为例,所提出模型取得了令人满意的结果:实际值与预测值的平均绝对误差(MAE)为0.00991mg/L,决定系数(R~2)达到了96.28%,平均绝对百分比误差(MAPE)为0.01647mg/L。与其他模型进行对比,本文提出的预测方法精度最高,验证了模型的精准性。实验结果表明所提模型具有较好的预测效果,能为水环境的污染治理提供有效的技术支持。更多还原

【Abstract】 To address the issue of low water quality prediction accuracy caused by significant and nonlinear fluctuations in water quality indicators,this paper proposes a hybrid water quality prediction model based on a Convolutional Neural Network and Bidirectional Gated Recurrent Unit(CNN-BiGRU) with a Multi-Head Attention Mechanism(MHA).Firstly,the correlation of various water quality indicators is calculated and filtered to select those significantly affecting water quality parameters as input variables.Secondly,the CNN-BiGRU hybrid model is constructed by combining the advantages of CNN in feature extraction and Bi-GRU networks in time series prediction,incorporating the Multi-Head Attention Mechanism(MHA).Finally,the feature matrix is input into the CNN-BiGRU-MHA model to achieve water quality prediction.Using publicly available national water quality data as a case study,and taking ammonia nitrogen prediction as an example,the proposed model yielded satisfactory results:the Mean Absolute Error(MAE) between actual and predicted values was0.00991mg/L,the Coefficient of Determination(R~2) reached 96.28%,and the Mean Absolute Percentage Error(MAPE) was 0.01647mg/L.Compared with other models,the proposed prediction method demonstrated the highest accuracy,validating the precision of the model.Experimental results indicate that the proposed model has excellent predictive performance,providing effective technical support for water environment pollution control.更多还原