基于长短期记忆神经网络的配电网负荷预测方法研究
|
摘要
高渗透率分布式光伏接入配电网后,将削减配电网负荷。由于光伏出力与配电网负荷均具有强随机性,且与温度、太阳辐照等相关气象因素耦合特性不同,导致配电网净负荷随机性提高、预测难度增加。为满足强波动性配电网净负荷短时预测需要,提出基于长短期记忆(long short term memory,LSTM)神经网络短期预测模型构建新方法。采用LSTM分别构建小时前配电网负荷预测模型和短期光伏出力预测模型,并分别使用交叉验证方法优化各个LSTM预测器结构超参数;最后,以两者预测结果相减,获得配电网净负荷。实测数据实验表明,相较于支持向量回归(SVR)等方法,采用LSTM的新方法能够自适应挖掘历史负荷、光伏出力特征与预测对象间的相关性,避免了复杂的特征选择环节,且预测精度优于SVR预测方法。
When high permeability distributed photovoltaic(DPV) is connected to the distribution network, the load of the distribution network will be reduced. As photovoltaic output and distribution network load have strong randomness and coupling characteristics with temperature, solar irradiance and other different related meteorological factors, the randomness of load in distribution network increased and difficult to forecast. In order to meet the need of short-term net load forecasting for distribution network with strong fluctuation, a new method of short-term load forecasting model based on long-term and shortterm memory(LSTM) neural network is proposed in this paper. Firstly, LSTM is used to construct load forecasting model of distribution network and short-term photovoltaic output forecasting model of one hour ago, then cross-validation method is used to optimize the Super-parameters of LSTM network for both of the predictors. Finally, the net load of distribution network is obtained by subtracting the forecasting results of the two methods. The experiment results based on measured data with the new method using LSTM can mine the relationship of load data and PV output between the predicting objects. Compared with traditional methods such as support vector regression(SVR), the new method has better prediction accuracy.
When high permeability distributed photovoltaic(DPV) is connected to the distribution network, the load of the distribution network will be reduced. As photovoltaic output and distribution network load have strong randomness and coupling characteristics with temperature, solar irradiance and other different related meteorological factors, the randomness of load in distribution network increased and difficult to forecast. In order to meet the need of short-term net load forecasting for distribution network with strong fluctuation, a new method of short-term load forecasting model based on long-term and shortterm memory(LSTM) neural network is proposed in this paper. Firstly, LSTM is used to construct load forecasting model of distribution network and short-term photovoltaic output forecasting model of one hour ago, then cross-validation method is used to optimize the Super-parameters of LSTM network for both of the predictors. Finally, the net load of distribution network is obtained by subtracting the forecasting results of the two methods. The experiment results based on measured data with the new method using LSTM can mine the relationship of load data and PV output between the predicting objects. Compared with traditional methods such as support vector regression(SVR), the new method has better prediction accuracy.
引文
[1]孙丰杰,谢宁,王承民,等.2017年国际供电会议配电系统规划研究成果综述[J].电网技术,2018,42(9):2733-2741.SUN Fengjie,XIE Ning,WANG Chengmin,et al.Review of CIRED 2017 on power distribution system planning[J].Power System Technology,2018,42(9):2733-2741.
[2]张立地,窦迅,王俊,等.综合能源背景下的配电网规划研究[J].供用电,2018,35(4):37-45.ZHANG Lidi,DOU Xun,WANG Jun,et al.Review of distribution network planning in the context of integrated energy[J].Distribution&Utilization,2018,35(4):37-45.
[3]曹煜祺,张立梅,白牧可.基于双层Elman神经网络的光伏发电功率预测[J].供用电,2017,34(10):8-13.CAO Yuqi,ZHANG Limei,BAI Muke.Photovoltaic power prediction based on double elman neural network[J].Distribution&Utilization,2017,34(10):8-13.
[4]李慧良,李鹏鹏,彭显刚,等.基于贝叶斯神经网络的短期负荷预测应用研究[J].广东电力,2012,25(11):16-19.LI Huiliang,LI Pengpeng,PENG Xiangang,et al.Short-term load prediction application research based on Bayes neural network[J].Guangdong Electric Power,2012,25(11):16-19.
[5]肖霖,张婧,曾鸣,等.基于小波分解和相同尺度序列的神经网络短期电价预测[J].电力需求侧管理,2011,13(4):19-22,29.XIAO Lin,ZHANG Jing,ZENG Ming,et al.Neural networks shortterm forecasting of electricity price based on wavelet decomposition and homo-layer series combination[J].Power Demand Side Management,2011,13(4):19-22,29.
[6]杨文佳,康重庆,夏清,等.基于预测误差分布特性统计分析的概率性短期负荷预测[J].电力系统自动化,2006,30(19):47-52.YANG Wenjia,KANG Chongqing,XIA Qing,et al.Short term probabilistic load forecasting based on statistics of probability distribution of forecasting errors[J].Automation of Electric Power Systems,2006,30(19):47-52.
[7]李东东,覃子珊,林顺富,等.基于混沌时间序列法的微网短期负荷预测[J].电力系统及其自动化学报,2015,27(5):14-18.LI Dongdong,QIN Zishan,LIN Shunfu,et al.Short-term load forecasting for microgrid based on method of chaotic time series[J].Proceedings of the CSU-EPSA,2015,27(5):14-18.
[8]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):67-72,92.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 Power Systems,2016,40(15):67-72,92.
[9]连立军,王艳君,邓林,等.基于多元统计分析的光伏发电量预测[J].河北农业大学学报,2017,40(1):111-116.LIAN Lijun,WANG Yanjun,DENG Lin,et al.Photovoltaic power generation prediction based on multivariate statistical analysis[J].Journal of Agricultural University of Hebei,2017,40(1):111-116.
[10]陈锦铭,郭雅娟,伍旺松,等.基于数据预处理与特征表示的多核SVM短期光伏发电预测[J].水电能源科学,2018,36(9):205-208,147.CHEN Jinming,GUO Yajuan,WU Wangsong,et al.Multi-kernel SVM short-term PV power prediction based on data pre-processing and characteristic representation[J].Water Resources and Power,2018,36(9):205-208,147.
[11]崔晓祥,李娟.基于支持向量机回归的电力系统负荷建模[J].江苏电机工程,2012,31(3):37-38,42.CUI Xiaoxiang,LI Juan.Load modeling based on SVM in power system[J].Jiangsu Electrical Engineering,2012,31(3):37-38,42.
[12]胡杨,常鲜戎.基于改进EMD-PSVM的短期负荷预测[J].陕西电力,2016,44(3):29-33.HU Yang, CHANG Xianrong.Short-term load forecasting based on improved EMD-PSVM[J].Shaanxi Electric Power,2016,44(3):29-33.
[13]耿博,高贞彦,白恒远,等.结合相似日GA-BP神经网络的光伏发电预测[J].电力系统及其自动化学报,2017,29(6):118-123.GENG Bo,GAO Zhenyan,BAI Hengyuan,et al.PV generation forecasting combined with similar days and GA-BP neural network[J].Proceedings of the CSU-EPSA,2017,29(6):118-123.
[14]霍娟,孙晓伟,张明杰.电力负荷预测算法比较-随机森林与支持向量机[J/OL].电力系统及其自动化学报:1-8[2019-05-13].https://doi.org/10.19635/j.cnki.csu-epsa.000093.HUO Juan,SUN Xiaowei,ZHANG Mingjie.Comparison of typical algorithms for load forecasting-random forest and support vector machine[J].Proceedings of the CSU-EPSA:1-8[2019-05-13].https://doi.org/10.19635/j.cnki.csu-epsa.000093.
[15]王守相,王亚旻,刘岩,等.基于经验模态分解和ELM神经网络的逐时太阳能辐照量预测[J].电力自动化设备,2014,34(8):7-12.WANG Shouxiang,WANG Yamin,LIU Yan,et al.Hourly solar radiation forecasting based on EMD and ELM neural network[J].Electric Power Automation Equipment,2014,34(8):7-12.
[16]弗朗索瓦·肖莱.Python深度学习[M].张亮,译.北京:人民邮电出版社,2018:6-17.
[17]林大贵.TensorFlow+Keras深度学习人工智能实践应用[M].北京:清华大学出版社,2018:193-196.
[18]李鹏,何帅,韩鹏飞,等.基于长短期记忆的实时电价条件下智能电网短期负荷预测[J].电网技术,2018,42(12):4045-4052.LI Peng,HE Shuai,HAN Pengfei,et al.Short-term load forecasting of smart grid based on long-short-term memory recurrent neural networks in condition of real-time electricity price[J].Power System Technology,2018,42(12):4045-4052.
[19]CHE J X,WANG J Z.Short-term load forecasting using a kernelbased support vector regression combination model[J].Applied Energy,2014,132:602-609.
[20]邱存勇,肖建.基于支持向量回归的电力系统短期负荷预测[J].计算机仿真,2013,30(11):62-65.QIU Cunyong,XIAO Jian.Power system short-term load forecasting based on support vector regression[J].Computer Simulation,2013,30(11):62-65.
[21]姚海成,周剑,林琳,等.利用蝙蝠算法优化SVR的太阳辐照度预测方法研究[J].可再生能源,2018,36(11):1612-1617.YAO Haicheng,ZHOU Jian,LIN Lin,et al.Solar radiation prediction method using bat algorithm optimized SVR[J].Renewable Energy Resources,2018,36(11):1612-1617.
[2]张立地,窦迅,王俊,等.综合能源背景下的配电网规划研究[J].供用电,2018,35(4):37-45.ZHANG Lidi,DOU Xun,WANG Jun,et al.Review of distribution network planning in the context of integrated energy[J].Distribution&Utilization,2018,35(4):37-45.
[3]曹煜祺,张立梅,白牧可.基于双层Elman神经网络的光伏发电功率预测[J].供用电,2017,34(10):8-13.CAO Yuqi,ZHANG Limei,BAI Muke.Photovoltaic power prediction based on double elman neural network[J].Distribution&Utilization,2017,34(10):8-13.
[4]李慧良,李鹏鹏,彭显刚,等.基于贝叶斯神经网络的短期负荷预测应用研究[J].广东电力,2012,25(11):16-19.LI Huiliang,LI Pengpeng,PENG Xiangang,et al.Short-term load prediction application research based on Bayes neural network[J].Guangdong Electric Power,2012,25(11):16-19.
[5]肖霖,张婧,曾鸣,等.基于小波分解和相同尺度序列的神经网络短期电价预测[J].电力需求侧管理,2011,13(4):19-22,29.XIAO Lin,ZHANG Jing,ZENG Ming,et al.Neural networks shortterm forecasting of electricity price based on wavelet decomposition and homo-layer series combination[J].Power Demand Side Management,2011,13(4):19-22,29.
[6]杨文佳,康重庆,夏清,等.基于预测误差分布特性统计分析的概率性短期负荷预测[J].电力系统自动化,2006,30(19):47-52.YANG Wenjia,KANG Chongqing,XIA Qing,et al.Short term probabilistic load forecasting based on statistics of probability distribution of forecasting errors[J].Automation of Electric Power Systems,2006,30(19):47-52.
[7]李东东,覃子珊,林顺富,等.基于混沌时间序列法的微网短期负荷预测[J].电力系统及其自动化学报,2015,27(5):14-18.LI Dongdong,QIN Zishan,LIN Shunfu,et al.Short-term load forecasting for microgrid based on method of chaotic time series[J].Proceedings of the CSU-EPSA,2015,27(5):14-18.
[8]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):67-72,92.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 Power Systems,2016,40(15):67-72,92.
[9]连立军,王艳君,邓林,等.基于多元统计分析的光伏发电量预测[J].河北农业大学学报,2017,40(1):111-116.LIAN Lijun,WANG Yanjun,DENG Lin,et al.Photovoltaic power generation prediction based on multivariate statistical analysis[J].Journal of Agricultural University of Hebei,2017,40(1):111-116.
[10]陈锦铭,郭雅娟,伍旺松,等.基于数据预处理与特征表示的多核SVM短期光伏发电预测[J].水电能源科学,2018,36(9):205-208,147.CHEN Jinming,GUO Yajuan,WU Wangsong,et al.Multi-kernel SVM short-term PV power prediction based on data pre-processing and characteristic representation[J].Water Resources and Power,2018,36(9):205-208,147.
[11]崔晓祥,李娟.基于支持向量机回归的电力系统负荷建模[J].江苏电机工程,2012,31(3):37-38,42.CUI Xiaoxiang,LI Juan.Load modeling based on SVM in power system[J].Jiangsu Electrical Engineering,2012,31(3):37-38,42.
[12]胡杨,常鲜戎.基于改进EMD-PSVM的短期负荷预测[J].陕西电力,2016,44(3):29-33.HU Yang, CHANG Xianrong.Short-term load forecasting based on improved EMD-PSVM[J].Shaanxi Electric Power,2016,44(3):29-33.
[13]耿博,高贞彦,白恒远,等.结合相似日GA-BP神经网络的光伏发电预测[J].电力系统及其自动化学报,2017,29(6):118-123.GENG Bo,GAO Zhenyan,BAI Hengyuan,et al.PV generation forecasting combined with similar days and GA-BP neural network[J].Proceedings of the CSU-EPSA,2017,29(6):118-123.
[14]霍娟,孙晓伟,张明杰.电力负荷预测算法比较-随机森林与支持向量机[J/OL].电力系统及其自动化学报:1-8[2019-05-13].https://doi.org/10.19635/j.cnki.csu-epsa.000093.HUO Juan,SUN Xiaowei,ZHANG Mingjie.Comparison of typical algorithms for load forecasting-random forest and support vector machine[J].Proceedings of the CSU-EPSA:1-8[2019-05-13].https://doi.org/10.19635/j.cnki.csu-epsa.000093.
[15]王守相,王亚旻,刘岩,等.基于经验模态分解和ELM神经网络的逐时太阳能辐照量预测[J].电力自动化设备,2014,34(8):7-12.WANG Shouxiang,WANG Yamin,LIU Yan,et al.Hourly solar radiation forecasting based on EMD and ELM neural network[J].Electric Power Automation Equipment,2014,34(8):7-12.
[16]弗朗索瓦·肖莱.Python深度学习[M].张亮,译.北京:人民邮电出版社,2018:6-17.
[17]林大贵.TensorFlow+Keras深度学习人工智能实践应用[M].北京:清华大学出版社,2018:193-196.
[18]李鹏,何帅,韩鹏飞,等.基于长短期记忆的实时电价条件下智能电网短期负荷预测[J].电网技术,2018,42(12):4045-4052.LI Peng,HE Shuai,HAN Pengfei,et al.Short-term load forecasting of smart grid based on long-short-term memory recurrent neural networks in condition of real-time electricity price[J].Power System Technology,2018,42(12):4045-4052.
[19]CHE J X,WANG J Z.Short-term load forecasting using a kernelbased support vector regression combination model[J].Applied Energy,2014,132:602-609.
[20]邱存勇,肖建.基于支持向量回归的电力系统短期负荷预测[J].计算机仿真,2013,30(11):62-65.QIU Cunyong,XIAO Jian.Power system short-term load forecasting based on support vector regression[J].Computer Simulation,2013,30(11):62-65.
[21]姚海成,周剑,林琳,等.利用蝙蝠算法优化SVR的太阳辐照度预测方法研究[J].可再生能源,2018,36(11):1612-1617.YAO Haicheng,ZHOU Jian,LIN Lin,et al.Solar radiation prediction method using bat algorithm optimized SVR[J].Renewable Energy Resources,2018,36(11):1612-1617.