基于RDPSO结构优化的三隐层BP神经网络水质预测模型及应用
|
摘要
提出基于随机漂移粒子群(RDPSO)算法结构优化的三隐层BP神经网络水质预测方法,利用RDPSO算法优化三隐层BP神经网络隐层节点数,构建三隐层RDPSO-BP水质预测模型,并与基于单隐层、双隐层节点优化的RDPSO-BP神经网络、基于网络权值阈值优化的RDPSO-BP神经网络预测模型作对比,以云南省五里冲水库总氮预测为例进行实例研究,利用实例60个月的总氮监测资料对此4种模型进行训练和预测。结果显示,三隐层RDPSO-BP神经网络模型对总氮预测的平均相对误差绝对值为6.98%,预测精度远高于其他3种模型,具有较好的预测精度和泛化能力。模型及方法可为相关水质预测研究提供参考。
A water quality prediction method of three-hidden layer BP neural network based on random drift particle swarm optimization( RDPSO) algorithm structure optimization is proposed. The RDPSO algorithm is used to optimize the number of hidden layer nodes of three-hidden layers BP neural network, and a three-hidden layer RDPSO-BP water quality prediction model is constructed. The models are compared with RDPSO-BP neural network based on single hidden layer and double-hidden layer node optimization, and the RDPSO-BP neural network based on network weight threshold optimization. The total nitrogen prediction of Wulichong Reservoir in Yunnan Province is taken as an example to study, and the total nitrogen monitoring data for 60 months in the example are used to carry out the four models. The results show that the absolute relative error of the three-hidden layer RDPSO-BP neural network model for total nitrogen prediction is 6. 98%, and the prediction accuracy is much higher than the other three models, which has better prediction accuracy and generalization ability. The models and methods can provide references for relevant water quality prediction studies.
A water quality prediction method of three-hidden layer BP neural network based on random drift particle swarm optimization( RDPSO) algorithm structure optimization is proposed. The RDPSO algorithm is used to optimize the number of hidden layer nodes of three-hidden layers BP neural network, and a three-hidden layer RDPSO-BP water quality prediction model is constructed. The models are compared with RDPSO-BP neural network based on single hidden layer and double-hidden layer node optimization, and the RDPSO-BP neural network based on network weight threshold optimization. The total nitrogen prediction of Wulichong Reservoir in Yunnan Province is taken as an example to study, and the total nitrogen monitoring data for 60 months in the example are used to carry out the four models. The results show that the absolute relative error of the three-hidden layer RDPSO-BP neural network model for total nitrogen prediction is 6. 98%, and the prediction accuracy is much higher than the other three models, which has better prediction accuracy and generalization ability. The models and methods can provide references for relevant water quality prediction studies.
引文
[1]张森,石为人,石欣,等.基于偏最小二乘回归和SVM的水质预测[J].计算机工程与应用,2015,51(15):249-254.
[2]高雅玉,张丽萍,闵祥宇,等.改进的BP神经网络在双塔水库水质预测中的应用[J].水资源与水工程学报,2012,23(6):149-153.
[3]王泽平.基于GA-BP与多隐层BP网络模型的水质预测及比较分析[J].水资源与水工程学报,2013,24(3):154-160.
[4]刘俊威,吕惠进.人工神经网络在水质预测中的应用研究[J].长江科学院院报,2012,29(9):95-97.
[5]程玉桂,黎明,林明玉.基于遗传算法和BP神经网络的城区中长期电力负荷预测与分析[J].计算机应用,2010,30(1):224-226.
[6]李松,罗勇,张铭锐.遗传算法优化BP神经网络的混沌时间序列预测[J].计算机工程与应用,2011,47(29):52-55.
[7]郭佩,何小海,陶青川,等.基于高维空间几何的PSO-BP神经网络图像复原[J].计算机工程与应用,2012,48(2):156-159.
[8]李松,刘力军,刘颖鹏.改进PSO优化BP神经网络的混沌时间序列预测[J].计算机工程与应用,2013,49(6):245-248.
[9]崔东文.多隐层BP神经网络模型在径流预测中的应用[J].水文,2013,33(1):68-73.
[10]查木哈,卢志宏,翟继武,等.双隐含层BP神经网络模型在老哈河水质预测中的应用[J].水资源与水工程学报,2018,29(2):56-61.
[11]韩爽,孟航,刘永前,等.增量处理双隐层BP神经网络风电功率预测模型[J].太阳能学报,2015,36(9):2238-2244.
[12]胡昌军.双隐层BP神经网络模型在区域水安全评价中的应用[J].水资源与水工程学报,2013,24(3):196-200.
[13]方伟,周建宏.基于频繁覆盖策略的随机漂移粒子群优化算法[J].控制与决策,2017,32(12):2127-2136.
[14]奚茂龙,卢丹,齐志明,等.随机漂移粒子群算法的RZWQM替代模型参数优化[J].计算机工程与应用,2016,52(23):35-41.
[15]刘安灵,黄振刚,王朝霞.基于改进随机漂移粒子群算法的微电网优化运行[J].电子设计工程,2018,26(6):33-37.
[16]张德丰,等.MATLAB神经网络应用设计[M].北京:机械工业出版社,2009.
[17]丛爽.面向MATLAB工具箱的神经网络理论与应用[M].合肥:中国科学技术大学出版社,2009.
[18]MATLAB中文论坛.MATLAB神经网络30个案例分析[M].北京:北京航空航天大学出版社,2010.
[2]高雅玉,张丽萍,闵祥宇,等.改进的BP神经网络在双塔水库水质预测中的应用[J].水资源与水工程学报,2012,23(6):149-153.
[3]王泽平.基于GA-BP与多隐层BP网络模型的水质预测及比较分析[J].水资源与水工程学报,2013,24(3):154-160.
[4]刘俊威,吕惠进.人工神经网络在水质预测中的应用研究[J].长江科学院院报,2012,29(9):95-97.
[5]程玉桂,黎明,林明玉.基于遗传算法和BP神经网络的城区中长期电力负荷预测与分析[J].计算机应用,2010,30(1):224-226.
[6]李松,罗勇,张铭锐.遗传算法优化BP神经网络的混沌时间序列预测[J].计算机工程与应用,2011,47(29):52-55.
[7]郭佩,何小海,陶青川,等.基于高维空间几何的PSO-BP神经网络图像复原[J].计算机工程与应用,2012,48(2):156-159.
[8]李松,刘力军,刘颖鹏.改进PSO优化BP神经网络的混沌时间序列预测[J].计算机工程与应用,2013,49(6):245-248.
[9]崔东文.多隐层BP神经网络模型在径流预测中的应用[J].水文,2013,33(1):68-73.
[10]查木哈,卢志宏,翟继武,等.双隐含层BP神经网络模型在老哈河水质预测中的应用[J].水资源与水工程学报,2018,29(2):56-61.
[11]韩爽,孟航,刘永前,等.增量处理双隐层BP神经网络风电功率预测模型[J].太阳能学报,2015,36(9):2238-2244.
[12]胡昌军.双隐层BP神经网络模型在区域水安全评价中的应用[J].水资源与水工程学报,2013,24(3):196-200.
[13]方伟,周建宏.基于频繁覆盖策略的随机漂移粒子群优化算法[J].控制与决策,2017,32(12):2127-2136.
[14]奚茂龙,卢丹,齐志明,等.随机漂移粒子群算法的RZWQM替代模型参数优化[J].计算机工程与应用,2016,52(23):35-41.
[15]刘安灵,黄振刚,王朝霞.基于改进随机漂移粒子群算法的微电网优化运行[J].电子设计工程,2018,26(6):33-37.
[16]张德丰,等.MATLAB神经网络应用设计[M].北京:机械工业出版社,2009.
[17]丛爽.面向MATLAB工具箱的神经网络理论与应用[M].合肥:中国科学技术大学出版社,2009.
[18]MATLAB中文论坛.MATLAB神经网络30个案例分析[M].北京:北京航空航天大学出版社,2010.