利用多源信息和深度置信神经网络的配电系统空间负荷预测
|
摘要
准确的空间负荷预测是配电系统精益化规划的基础。在此背景下,提出利用多源信息融合和深度置信神经网络的配电系统空间负荷预测方法。首先,在分析空间负荷元胞多源信息特征的基础上,采用基于程度副词语义标定的结构化方法对负荷元胞的非结构化属性进行结构化处理,以充分挖掘利用负荷元胞数据信息。然后,采用受限玻尔兹曼机方法和反向传播(back propagation,BP)算法相结合学习元胞特征,以提升元胞高维特征提取的性能,并采用训练后的深度置信神经网络预测待规划区域的空间饱和负荷密度。最后,以某城市的区域配电系统为例,对所提出的空间负荷预测方法进行验证;仿真结果表明:在空间负荷预测模型中考虑非结构化信息的影响可以提高空间负荷预测精度,且与现有的一些方法相比,所提方法的预测精度更高。
Accurate spatial load forecasting is of great significance for promoting fine planning of distribution systems. A spatial electric load forecasting method for distribution systems is proposed by using multi-source information and the deep belief netw ork( DBN) and deep neural network( DNN)( DBN-DNN). First,the multi-source information feature of cell loads is analyzed,and then a structured method based on the quantification of degree adverb is utilized to transform the unstructured attributes for digging and using the data information of cell loads fully. Then,both the restricted Boltzmann machine( RBM) method and back propagation( BP) algorithm based feedforw ard neural netw ork are adopted to learn cellular features for enhancing the performance of extracting high-dimensional features of cell loads,and the spatial saturation load density of the planning area is forecasted by the trained DBN-DNN model. Finally,the distribution system in a part of a city is employed for demonstrating the effectiveness of the proposed spatial load forecasting method. Numerical results demonstrated that more accurate spatial load forecasting results can be obtained with the proposed method by considering unstructured attributes of cell loads or comparing with the some existing methods.
Accurate spatial load forecasting is of great significance for promoting fine planning of distribution systems. A spatial electric load forecasting method for distribution systems is proposed by using multi-source information and the deep belief netw ork( DBN) and deep neural network( DNN)( DBN-DNN). First,the multi-source information feature of cell loads is analyzed,and then a structured method based on the quantification of degree adverb is utilized to transform the unstructured attributes for digging and using the data information of cell loads fully. Then,both the restricted Boltzmann machine( RBM) method and back propagation( BP) algorithm based feedforw ard neural netw ork are adopted to learn cellular features for enhancing the performance of extracting high-dimensional features of cell loads,and the spatial saturation load density of the planning area is forecasted by the trained DBN-DNN model. Finally,the distribution system in a part of a city is employed for demonstrating the effectiveness of the proposed spatial load forecasting method. Numerical results demonstrated that more accurate spatial load forecasting results can be obtained with the proposed method by considering unstructured attributes of cell loads or comparing with the some existing methods.
引文
[1]肖白,周潮,穆钢.空间电力负荷预测方法综述与展望[J].中国电机工程学报,2013,33(25):78-92.XIAO Bai,ZHOU Chao,M U Gang. Review and prospect of the spatial load forecasting methods[J]. Proceedings of the CSEE,2013,33(25):78-92.
[2]刘琪琛,雷景生,郝珈玮,等.基于Spark平台和并行随机森林回归算法的短期电力负荷预测[J].电力建设,2017,38(10):84-92.LIU Qichen,LEI Jingsheng,HAO Jiaw ei,et al. Short-term pow er load forecasting based on Spark platform and parallel random forest regression algorithm model[J]. Electric Pow er Construction,2017,38(10):84-92.
[3]雷景生,郝珈玮,朱国康.基于“分层-汇集”模型的短期电力负荷预测[J].电力建设,2017,38(1):68-75.LEI Jingsheng,HAO Jiaw ei,ZHU Guokang. Short-term pow er load forecasting based on ‘layered-confluence’ model[J]. Electric Pow er Construction,2017,38(1):68-75.
[4]杨军胜,彭石,王承民,等.基于城市用地性质的配网空间负荷预测研究[J].电测与仪表,2018,55(11):30-34.YANG Junsheng,PENG Shi,WANG Chengmin,et al. Research of spatial load forecasting in distribution netw ork based on grid partition of urban land-use property[J]. Electrical M easurement&Instrumentation,2018,55(11):30-34.
[5]肖白,杨修宇,穆钢,等.基于元胞历史负荷数据的负荷密度指标法[J].电网技术,2014,38(4):1014-1019.XIAO Bai,YANG Xiuyu,M U Gang,et al. A load density index method based on historical data of cell load[J]. Pow er System Technology,2014,38(4):1014-1019.
[6]肖白,杨欣桐,田莉,等.计及元胞发展程度的空间负荷预测方法[J].电力系统自动化,2018,42(1):61-67.XIAO Bai, YANG Xintong, TIAN Li, et al. Spatial load forecasting method based on development degree of cell[J].Automation of Electric Pow er Systems,2018,42(1):61-67.
[7]肖白,刘庆永,牛强,等.基于元胞负荷特性分析的RBF神经网络空间负荷预测方法[J].电网技术,2018,42(1):301-307.XIAO Bai, LIU qingyong, NIU Qiang, et al. A spatial load forecasting method based on RBF neural netw ork and cellular load characteristics analysis[J]. Pow er System Technology,2018,42(1):301-307.
[8]周湶,孙威,张昀,等.基于改进型ANFIS的负荷密度指标求取新方法[J].电力系统保护与控制,2011,39(1):29-34.ZHOU Quan,SUN Wei,ZHANG Yun,et al. A new method to obtain load density based on improved ANFIS[J]. Pow er System Protection and Control,2011,39(1):29-34.
[9]周湶,孙威,任海军,等.基于最小二乘支持向量机和负荷密度指标法的配电网空间负荷预测[J].电网技术,2011,35(1):66-71.ZHOU Quan, SUN Wei, REN Haijun, et al. Spatial load forecasting of distribution netw ork based on least squares support vector machine and load density index system[J]. Pow er System Technology,2011,35(1):66-71.
[10]刘思,傅旭华,叶承晋,等.应用聚类分析与非参数核密度估计的空间负荷分布规律[J].电网技术,2017,41(2):604-609.LIU Si,FU Xuhua,YE Chengjin,et al. Spatial load distribution based on clustering analysis and non-parametric kernel density estimation[J]. Pow er System Technology, 2017, 41(2):604-609.
[11]刘自发,庞铖铖,王泽黎,等.基于云理论和元胞自动机理论的城市配电网空间负荷预测[J].中国电机工程学报,2013,33(10):98-105.LIU Zifa,PANG Chengcheng,WANG Zeli,et al. Spatial load forecasting for distribution netw ork based on cloud theory andcellular automata[J]. Proceedings of the CSEE,2013,33(10):98-105.
[12]吴倩红,高军,侯广松,等.实现影响因素多源异构融合的短期负荷预测支持向量机算法[J].电力系统自动化,2016,40(15):67-72.WU Qianhong,GAO Jun,HOU Guangsong,et al. Short-term load forecasting support vector machine algorithm based on multi-source heterogeneous fusion of load factors[J]. Automation of Electric Pow er Systems,2016,40(15):67-72.
[13]朱凤娟,王主丁,陆俭,等.考虑小区发展不均衡的空间负荷预测分类分区法[J].电力系统自动化,2012,36(12):41-48.ZHU Fengjuan,WANG Zhuding,LU Jian,et al. Disequilibrium development areas based on classification and subarea method for spatial load forecasting[J]. Automation of Electric Pow er Systems,2012,36(12):41-48.
[14]张帆,刘杰锋,李冰,等.基于改进Logistic模型阶段划分理论的饱和负荷预测[J].电力建设,2015,36(10):105-110.ZHANG Fan, LIU Jiefeng, LI Bing, et al. Saturated load forecasting based on stages division theory of improved Logistic model[J]. Electric Pow er Construction,2015,36(10):105-110.
[15]肖白,刘庆永,房龙江,等.基于模糊粗糙集理论和时空信息的空间负荷预测[J].电力建设,2017,38(1):58-67.XIAO Bai,LIU Qingyong,FANG Longjiang,et al. Spatial load forecasting based on fuzzy rough set theory w ith spatial and temporal information[J]. Electric Pow er Construction,2017,38(1):58-67.
[16]吉兴全,傅荣荣,文福拴,等.饱和负荷预测中的多级聚类分析和改进Logistic模型[J].电力系统及其自动化学报,2017,29(8):138-144.JI Xingquan, FU Rongrong, WEN Fushuan, et al. M ultilevel clustering analysis and improved Logistic model for saturated load forecasting[J]. Proceedings of the CSU-EPSA,2017,29(8):138-144.
[17]申超群.基于负荷密度法与Logistic模型法的饱和负荷预测研究[D].郑州:郑州大学,2016.SHEN Chaoqun. Research on saturated load forecasting based on load density method and logistic model method[D]. Zhengzhou:Zhengzhou University,2016.
[18]刘杰锋,程浩忠,韩新阳,等.多维度饱和负荷预测方法及其应用[J].电力系统及其自动化学报,2015,27(2):44-50.LIU Jiefeng, CHENG Haozhong, HAN Xinyang, et al.Forecasting method and application of multi-dimensional saturated load[J]. Proceedings of the CSU-EPSA,2015,27(2):44-50.
[19]马润泽.基于大数据技术的配电网规划研究[D].杭州:浙江大学,2018.M A Runze. Research on distribution netw ork planning based on big data technology[D]. Hangzhou:Zhejiang University,2018.
[20]马润泽,王龙响,余佳文,等.考虑历史缺陷文本信息的断路器状态评价研究[J].机电工程,2015,32(10):1375-1379.M A Runze, WANG Longxiang, YU Jiaw en, et al. Circuit breakers’ condition evaluation considering the information in historical defect texts[J]. Journal of M echanical&Electrical Engineering,2015,32(10):1375-1379.
[21]刘梓权,王慧芳,曹靖,等.基于卷积神经网络的电力设备缺陷文本分类模型研究[J].电网技术,2018,42(2):644-650.LIU Ziquan,WANG Huifang,CAO Jing,et al. A classification model of pow er equipment defect texts based on convolutional neural netw ork[J]. Pow er System Technology,2018,42(2):644-650.
[22]何永强,秦勤,王俊鹏.基于深度神经网络的嵌入式向量及话题模型[J].计算机工程与设计,2016,37(12):3384-3388.HE Yongqiang,QIN Qing,WANG Junpeng. Embedded vector and topic model based on deep neural netw orks[J]. Computer Engineering and Design,2016,37(12):3384-3388.
[23] MA X,WANG H,GENG J. Spectral–spatial classification of hyperspectral image based on deep auto-encoder[J]. IEEE Journal of Selected Topics in Applied Earth Observations&Remote Sensing,2016,9(9):4073-4085.
[24]其米克·巴特西,黄浩,王羡慧.基于深度神经网络的维吾尔语语音识别[J].计算机工程与设计,2015,36(8):2239-2244.QIM IKE Batexi,HUANG Hao,WANG Xianhui. Uyghur speech recognition based on deep neural netw ork[J]. Computer Engineering and Design,2015,36(8):2239-2244.
[25] HINTON G E, SALAKHUTDINOV R R. Reducing the dimensionality of data w ith neural netw orks[J]. Science,2006,313(5786):504-507.
[26]深度学习-深度信念(置信)网络(DBN)-从原理到实现(DeepLearnToolBox)[EB/OL].[2018-08-12]. https://blog.csdn. net/Rainbow0210/article/details/53010694.
[27]深度置信神经网络(DBN-DNN)[EB/OL].[2018-08-12].https://blog. csdn. net/zhanglu_w ind/article/details/78949020.
[28] LI J,ZHANG T,LUO W,et al. Sparseness analysis in the pretraining of deep neural netw orks[J]. IEEE Transactions on Neural Netw orks and Learning Systems,2016,28(6):1425-1438.
[29] MA Runze,YU Peng,HUANG Minxiang. Spatial load forecasting based on unstructured information processing and multi-attribute deep learning[C]//Proceedings of the 5th International Symposium on Electrical and Electronics Engineering(ISEEE),Galati,Romania. IEEE,2017:1-7.
[2]刘琪琛,雷景生,郝珈玮,等.基于Spark平台和并行随机森林回归算法的短期电力负荷预测[J].电力建设,2017,38(10):84-92.LIU Qichen,LEI Jingsheng,HAO Jiaw ei,et al. Short-term pow er load forecasting based on Spark platform and parallel random forest regression algorithm model[J]. Electric Pow er Construction,2017,38(10):84-92.
[3]雷景生,郝珈玮,朱国康.基于“分层-汇集”模型的短期电力负荷预测[J].电力建设,2017,38(1):68-75.LEI Jingsheng,HAO Jiaw ei,ZHU Guokang. Short-term pow er load forecasting based on ‘layered-confluence’ model[J]. Electric Pow er Construction,2017,38(1):68-75.
[4]杨军胜,彭石,王承民,等.基于城市用地性质的配网空间负荷预测研究[J].电测与仪表,2018,55(11):30-34.YANG Junsheng,PENG Shi,WANG Chengmin,et al. Research of spatial load forecasting in distribution netw ork based on grid partition of urban land-use property[J]. Electrical M easurement&Instrumentation,2018,55(11):30-34.
[5]肖白,杨修宇,穆钢,等.基于元胞历史负荷数据的负荷密度指标法[J].电网技术,2014,38(4):1014-1019.XIAO Bai,YANG Xiuyu,M U Gang,et al. A load density index method based on historical data of cell load[J]. Pow er System Technology,2014,38(4):1014-1019.
[6]肖白,杨欣桐,田莉,等.计及元胞发展程度的空间负荷预测方法[J].电力系统自动化,2018,42(1):61-67.XIAO Bai, YANG Xintong, TIAN Li, et al. Spatial load forecasting method based on development degree of cell[J].Automation of Electric Pow er Systems,2018,42(1):61-67.
[7]肖白,刘庆永,牛强,等.基于元胞负荷特性分析的RBF神经网络空间负荷预测方法[J].电网技术,2018,42(1):301-307.XIAO Bai, LIU qingyong, NIU Qiang, et al. A spatial load forecasting method based on RBF neural netw ork and cellular load characteristics analysis[J]. Pow er System Technology,2018,42(1):301-307.
[8]周湶,孙威,张昀,等.基于改进型ANFIS的负荷密度指标求取新方法[J].电力系统保护与控制,2011,39(1):29-34.ZHOU Quan,SUN Wei,ZHANG Yun,et al. A new method to obtain load density based on improved ANFIS[J]. Pow er System Protection and Control,2011,39(1):29-34.
[9]周湶,孙威,任海军,等.基于最小二乘支持向量机和负荷密度指标法的配电网空间负荷预测[J].电网技术,2011,35(1):66-71.ZHOU Quan, SUN Wei, REN Haijun, et al. Spatial load forecasting of distribution netw ork based on least squares support vector machine and load density index system[J]. Pow er System Technology,2011,35(1):66-71.
[10]刘思,傅旭华,叶承晋,等.应用聚类分析与非参数核密度估计的空间负荷分布规律[J].电网技术,2017,41(2):604-609.LIU Si,FU Xuhua,YE Chengjin,et al. Spatial load distribution based on clustering analysis and non-parametric kernel density estimation[J]. Pow er System Technology, 2017, 41(2):604-609.
[11]刘自发,庞铖铖,王泽黎,等.基于云理论和元胞自动机理论的城市配电网空间负荷预测[J].中国电机工程学报,2013,33(10):98-105.LIU Zifa,PANG Chengcheng,WANG Zeli,et al. Spatial load forecasting for distribution netw ork based on cloud theory andcellular automata[J]. Proceedings of the CSEE,2013,33(10):98-105.
[12]吴倩红,高军,侯广松,等.实现影响因素多源异构融合的短期负荷预测支持向量机算法[J].电力系统自动化,2016,40(15):67-72.WU Qianhong,GAO Jun,HOU Guangsong,et al. Short-term load forecasting support vector machine algorithm based on multi-source heterogeneous fusion of load factors[J]. Automation of Electric Pow er Systems,2016,40(15):67-72.
[13]朱凤娟,王主丁,陆俭,等.考虑小区发展不均衡的空间负荷预测分类分区法[J].电力系统自动化,2012,36(12):41-48.ZHU Fengjuan,WANG Zhuding,LU Jian,et al. Disequilibrium development areas based on classification and subarea method for spatial load forecasting[J]. Automation of Electric Pow er Systems,2012,36(12):41-48.
[14]张帆,刘杰锋,李冰,等.基于改进Logistic模型阶段划分理论的饱和负荷预测[J].电力建设,2015,36(10):105-110.ZHANG Fan, LIU Jiefeng, LI Bing, et al. Saturated load forecasting based on stages division theory of improved Logistic model[J]. Electric Pow er Construction,2015,36(10):105-110.
[15]肖白,刘庆永,房龙江,等.基于模糊粗糙集理论和时空信息的空间负荷预测[J].电力建设,2017,38(1):58-67.XIAO Bai,LIU Qingyong,FANG Longjiang,et al. Spatial load forecasting based on fuzzy rough set theory w ith spatial and temporal information[J]. Electric Pow er Construction,2017,38(1):58-67.
[16]吉兴全,傅荣荣,文福拴,等.饱和负荷预测中的多级聚类分析和改进Logistic模型[J].电力系统及其自动化学报,2017,29(8):138-144.JI Xingquan, FU Rongrong, WEN Fushuan, et al. M ultilevel clustering analysis and improved Logistic model for saturated load forecasting[J]. Proceedings of the CSU-EPSA,2017,29(8):138-144.
[17]申超群.基于负荷密度法与Logistic模型法的饱和负荷预测研究[D].郑州:郑州大学,2016.SHEN Chaoqun. Research on saturated load forecasting based on load density method and logistic model method[D]. Zhengzhou:Zhengzhou University,2016.
[18]刘杰锋,程浩忠,韩新阳,等.多维度饱和负荷预测方法及其应用[J].电力系统及其自动化学报,2015,27(2):44-50.LIU Jiefeng, CHENG Haozhong, HAN Xinyang, et al.Forecasting method and application of multi-dimensional saturated load[J]. Proceedings of the CSU-EPSA,2015,27(2):44-50.
[19]马润泽.基于大数据技术的配电网规划研究[D].杭州:浙江大学,2018.M A Runze. Research on distribution netw ork planning based on big data technology[D]. Hangzhou:Zhejiang University,2018.
[20]马润泽,王龙响,余佳文,等.考虑历史缺陷文本信息的断路器状态评价研究[J].机电工程,2015,32(10):1375-1379.M A Runze, WANG Longxiang, YU Jiaw en, et al. Circuit breakers’ condition evaluation considering the information in historical defect texts[J]. Journal of M echanical&Electrical Engineering,2015,32(10):1375-1379.
[21]刘梓权,王慧芳,曹靖,等.基于卷积神经网络的电力设备缺陷文本分类模型研究[J].电网技术,2018,42(2):644-650.LIU Ziquan,WANG Huifang,CAO Jing,et al. A classification model of pow er equipment defect texts based on convolutional neural netw ork[J]. Pow er System Technology,2018,42(2):644-650.
[22]何永强,秦勤,王俊鹏.基于深度神经网络的嵌入式向量及话题模型[J].计算机工程与设计,2016,37(12):3384-3388.HE Yongqiang,QIN Qing,WANG Junpeng. Embedded vector and topic model based on deep neural netw orks[J]. Computer Engineering and Design,2016,37(12):3384-3388.
[23] MA X,WANG H,GENG J. Spectral–spatial classification of hyperspectral image based on deep auto-encoder[J]. IEEE Journal of Selected Topics in Applied Earth Observations&Remote Sensing,2016,9(9):4073-4085.
[24]其米克·巴特西,黄浩,王羡慧.基于深度神经网络的维吾尔语语音识别[J].计算机工程与设计,2015,36(8):2239-2244.QIM IKE Batexi,HUANG Hao,WANG Xianhui. Uyghur speech recognition based on deep neural netw ork[J]. Computer Engineering and Design,2015,36(8):2239-2244.
[25] HINTON G E, SALAKHUTDINOV R R. Reducing the dimensionality of data w ith neural netw orks[J]. Science,2006,313(5786):504-507.
[26]深度学习-深度信念(置信)网络(DBN)-从原理到实现(DeepLearnToolBox)[EB/OL].[2018-08-12]. https://blog.csdn. net/Rainbow0210/article/details/53010694.
[27]深度置信神经网络(DBN-DNN)[EB/OL].[2018-08-12].https://blog. csdn. net/zhanglu_w ind/article/details/78949020.
[28] LI J,ZHANG T,LUO W,et al. Sparseness analysis in the pretraining of deep neural netw orks[J]. IEEE Transactions on Neural Netw orks and Learning Systems,2016,28(6):1425-1438.
[29] MA Runze,YU Peng,HUANG Minxiang. Spatial load forecasting based on unstructured information processing and multi-attribute deep learning[C]//Proceedings of the 5th International Symposium on Electrical and Electronics Engineering(ISEEE),Galati,Romania. IEEE,2017:1-7.