基于广义回归神经网络的日总辐射曝辐量预估
|
摘要
采用广义回归神经网络(generalized regression neural network,GRNN)模型对山东烟台市福山气象站2000～2003年日总辐射曝辐量进行预估。模型通过交叉验证方法确定其关键参数(光滑因子),以日照时数、平均气压、平均气温、日最高气温、相对湿度、气溶胶光学厚度6个变量作为输入量。结果显示:GRNN15.9%,均方根误差为2.32 MJ/m2,拟合优度为0.892,且模型的预估精度和拟合优度均明显优于LM-BP网络。气溶胶光学厚度对GRNN是预估当地日总辐射曝辐量的一种有效方法。
The generalized regression neural network(GRNN)model is used to estimate the daily irradiation exposure of global radiation from 2000 to 2003 at Fushan Meteorological Station,Yantai city,Shandong province. The cross validation method is adopted to determine the key parameter of GRNN model(smooth factor). The input parameters of GRNN model included sunshine duration,average pressure,average air temperature,daily maximum air temperature,relative humidity,and aerosol optical thickness at Fushan meteorological station. Results are promising with MPE is 15.9%,RMSE is 2.32 MJ/m~2,and the correlation coefficient is 0.892. The optimized GRNN presents the estimate better than LM-BP network. Aerosol optical thickness in the model has almost no influence on both MPE and RMSE. Therefore,using the GRNN model with the meteorological observation data is a very effective method to estimate the daily irradiation exposure of global radiation of some region which has not radiation observation site.
The generalized regression neural network(GRNN)model is used to estimate the daily irradiation exposure of global radiation from 2000 to 2003 at Fushan Meteorological Station,Yantai city,Shandong province. The cross validation method is adopted to determine the key parameter of GRNN model(smooth factor). The input parameters of GRNN model included sunshine duration,average pressure,average air temperature,daily maximum air temperature,relative humidity,and aerosol optical thickness at Fushan meteorological station. Results are promising with MPE is 15.9%,RMSE is 2.32 MJ/m~2,and the correlation coefficient is 0.892. The optimized GRNN presents the estimate better than LM-BP network. Aerosol optical thickness in the model has almost no influence on both MPE and RMSE. Therefore,using the GRNN model with the meteorological observation data is a very effective method to estimate the daily irradiation exposure of global radiation of some region which has not radiation observation site.
引文
[1]朱良山,林婵,王起峰. LM_BP神经网络在总辐射日曝辐量计算中的应用研究[J].太阳能学报,2013,34(7):1202—1205.[1] Zhu Liangshan,Lin Chan,Wang Qifeng. Applicationsof LM_BP network in daily irradiation exposure of globalradiation calculate[J]. Acta Energiae Solaris Sinica,2013,34(7):1202—1205.
[2]周晋,吴业正,晏刚,等.利用神经网络估算太阳辐射[J].太阳能学报,2005,24(4):509—512.[2] Zhou Jin, Wu Yezheng, Yan Gang, et al. Solarradiation estimation using artificial neural networks[J].Acta Energiae Solaris Sinica,2005,24(4):509—512.
[3]李净,王卫东,张福存.利用LM-BP神经网络估算西北地区太阳辐射[J].干旱区地理,2015,38(3):439—445.[3] Li Jing,Wang Weidong,Zhang Fucun. Simulation ofsolar radiation in northwest China based on LM-BPneural network[J]. Arid Land Geography,2015,38(3):439—445.
[4]伊良忠,章超,裴峥.广义回归神经网络的改进及在交通预测中的应用[J].山东大学学报:工学版,2013,43(1):9—14.[4] Yi Liangzhong,Zhang Chao,Pei Zheng. A modifiedgeneral regression neural network with its application intraffic prediction[J]. Journal of Shandong University:Engineering Science,2013,43(1):9—14.
[5]郑含博.电力变压器状态评估及故障诊断方法研究[D].重庆:重庆大学电气工程学院,2012.[5] Zheng Hanbo. Study on condition assessment and faultdiagnosis approaches for power transformers[D].Chongqing:College of Electrical Engineering ofChongqing University,2012.
[6]周建萍,闫澍旺.广义回归神经网络预测加筋土支挡结构高度[J].岩土力学,2002,23(4):486—490.[6] Zhou Jianping,Yan Shuwang. Generalized regressionneural networks for predicting design height of GRW[J]. Rock and Soil Mechanics,2002,23(4):486—490.
[7]冯志鹏,宋希庚,薛冬新.基于广义回归神经网络的时间序列预测研究[J].振动、测试与诊断,2003,23(2):105—109.[7] Feng Zhipeng,Song Xigeng,Xue Dongxin. Generalregression neural network based prediction of time series[J]. Journal of Vibration,Measurement&Diagnosis,2003,23(2):105—109.
[8]卓莹,张强,龚正虎.网络态势预测的广义回归神经网络模型[J].解放军理工大学学报:自然科学版,2012,13(2):148—151.[8] Zhuo Ying,Zhang Qiang,Gong Zhenghu. GRNN modelof network situation forecast[J]. Journal of PLAUniversity of Science and Technology:Natural ScienceEdition,2012,13(2):148—151.
[9]王菲露,宋杨.基于广义回归神经网络的蛋白质二级结构预测[J].计算机仿真,2012,29(2):184—187.[9] Wang Feilu,Song Yang. Protein secondary structureprediction based on generalized regression neuralnetwork[J]. Computer Simulation,2012,29(2):184—187.
[10]张德丰. MATLAB神经网络应用设计[M].北京:机械工业出版社,2009.[10] Zhang Defeng. Application design of MATLAB neuralnetwork[M]. Beijing:Machanical Industry Press,2009.
[11]董长虹. MATLAB神经网络与应用[M].北京:国防工业出版社,2005.[11] Dong Changhong. Neural network and application ofMATLAB[M]. Beijing:National Defense IndusteyPress,2005.
[12] 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.[12] Matlab Chinese BBS. 30 cases analysis of MATLABneural network[M]. Beijing:Beijing Universty ofAeronautics and Astronautics Press,2010.
[13]邱龙金,贺昌政.神经网络稳定性的交叉验证模型[J].计算机工程与应用,2010,46(34):43—45.[13] Qiu Longjin,He Changzheng. Cross validation model forneural network stability[J]. Computer Engineering andApplications,2010,46(34):43—45.
[2]周晋,吴业正,晏刚,等.利用神经网络估算太阳辐射[J].太阳能学报,2005,24(4):509—512.[2] Zhou Jin, Wu Yezheng, Yan Gang, et al. Solarradiation estimation using artificial neural networks[J].Acta Energiae Solaris Sinica,2005,24(4):509—512.
[3]李净,王卫东,张福存.利用LM-BP神经网络估算西北地区太阳辐射[J].干旱区地理,2015,38(3):439—445.[3] Li Jing,Wang Weidong,Zhang Fucun. Simulation ofsolar radiation in northwest China based on LM-BPneural network[J]. Arid Land Geography,2015,38(3):439—445.
[4]伊良忠,章超,裴峥.广义回归神经网络的改进及在交通预测中的应用[J].山东大学学报:工学版,2013,43(1):9—14.[4] Yi Liangzhong,Zhang Chao,Pei Zheng. A modifiedgeneral regression neural network with its application intraffic prediction[J]. Journal of Shandong University:Engineering Science,2013,43(1):9—14.
[5]郑含博.电力变压器状态评估及故障诊断方法研究[D].重庆:重庆大学电气工程学院,2012.[5] Zheng Hanbo. Study on condition assessment and faultdiagnosis approaches for power transformers[D].Chongqing:College of Electrical Engineering ofChongqing University,2012.
[6]周建萍,闫澍旺.广义回归神经网络预测加筋土支挡结构高度[J].岩土力学,2002,23(4):486—490.[6] Zhou Jianping,Yan Shuwang. Generalized regressionneural networks for predicting design height of GRW[J]. Rock and Soil Mechanics,2002,23(4):486—490.
[7]冯志鹏,宋希庚,薛冬新.基于广义回归神经网络的时间序列预测研究[J].振动、测试与诊断,2003,23(2):105—109.[7] Feng Zhipeng,Song Xigeng,Xue Dongxin. Generalregression neural network based prediction of time series[J]. Journal of Vibration,Measurement&Diagnosis,2003,23(2):105—109.
[8]卓莹,张强,龚正虎.网络态势预测的广义回归神经网络模型[J].解放军理工大学学报:自然科学版,2012,13(2):148—151.[8] Zhuo Ying,Zhang Qiang,Gong Zhenghu. GRNN modelof network situation forecast[J]. Journal of PLAUniversity of Science and Technology:Natural ScienceEdition,2012,13(2):148—151.
[9]王菲露,宋杨.基于广义回归神经网络的蛋白质二级结构预测[J].计算机仿真,2012,29(2):184—187.[9] Wang Feilu,Song Yang. Protein secondary structureprediction based on generalized regression neuralnetwork[J]. Computer Simulation,2012,29(2):184—187.
[10]张德丰. MATLAB神经网络应用设计[M].北京:机械工业出版社,2009.[10] Zhang Defeng. Application design of MATLAB neuralnetwork[M]. Beijing:Machanical Industry Press,2009.
[11]董长虹. MATLAB神经网络与应用[M].北京:国防工业出版社,2005.[11] Dong Changhong. Neural network and application ofMATLAB[M]. Beijing:National Defense IndusteyPress,2005.
[12] 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.[12] Matlab Chinese BBS. 30 cases analysis of MATLABneural network[M]. Beijing:Beijing Universty ofAeronautics and Astronautics Press,2010.
[13]邱龙金,贺昌政.神经网络稳定性的交叉验证模型[J].计算机工程与应用,2010,46(34):43—45.[13] Qiu Longjin,He Changzheng. Cross validation model forneural network stability[J]. Computer Engineering andApplications,2010,46(34):43—45.