考虑优化ARIMA模型差分次数的风功率预测
|
摘要
针对现有差分自回归移动平均模型进行风功率预测不具有普遍适用性问题,对差分自回归移动平均模型进行改进,提出了一种确定不同出力特性的风电场风功率时间序列转化为平稳序列所需的最优差分次数的方法。应用增广迪基-福勒检验判断序列的平稳性,分别以赤池信息准则、Yule-Walker方程以及移动平均参数和自协方差方程的关系确定出模型阶数、自回归参数、移动平均参数,并加入限幅环节对预测结果进行修正。以昌图风电场的原始出力数据为例,以图形的形式直观分析了原始风电出力序列的概率分布特性、时间相关性、时间分布特性和波动特性等性质,验证了预测序列满足原序列的性质。以误差、均方差、平均绝对误差为预测评价指标,与原差分自回归移动平均模型相比,所提出的改进差分自回归移动平均模型具有更好的预测效果。
Considering that the existing autoregressive integrated moving average(ARIMA)models do not have universal applicability in wind power forecast,an improved ARIMA model is proposed,which focuses on determining the optimal differential times of wind power time series with various output characteristics to be converted into a stationary series. The stationarity of wind power series is validated using augmented Dickey-Fuller(ADF)test. The order,autoregressive parameters and moving average parameters are determined by Akaike's information criterion(AIC)criterion,Yule-Walker equation and the relationship between moving average parameters and auto-covariance equation,respectively;moreover,the forecasting results are modified by the addition of limiting link. With the original output data from Changtu wind farm as an example,the probability distribution,temporal correlation,time distribution and fluctuation characteristics of the original wind power generation series are analyzed in the form of graphics,which verifies that the forecasting series can satisfy the characteristics of the original series. Compared with the original ARIMA model,the improved ARIMA model proposed in this paper has better performance according to the evaluation indexes for forecasting,such as error,mean square error and mean absolute error.
Considering that the existing autoregressive integrated moving average(ARIMA)models do not have universal applicability in wind power forecast,an improved ARIMA model is proposed,which focuses on determining the optimal differential times of wind power time series with various output characteristics to be converted into a stationary series. The stationarity of wind power series is validated using augmented Dickey-Fuller(ADF)test. The order,autoregressive parameters and moving average parameters are determined by Akaike's information criterion(AIC)criterion,Yule-Walker equation and the relationship between moving average parameters and auto-covariance equation,respectively;moreover,the forecasting results are modified by the addition of limiting link. With the original output data from Changtu wind farm as an example,the probability distribution,temporal correlation,time distribution and fluctuation characteristics of the original wind power generation series are analyzed in the form of graphics,which verifies that the forecasting series can satisfy the characteristics of the original series. Compared with the original ARIMA model,the improved ARIMA model proposed in this paper has better performance according to the evaluation indexes for forecasting,such as error,mean square error and mean absolute error.
引文
[1]崔杨,穆钢,刘玉,等(Cui Yang,Mu Gang,Liu Yu,et al).风电功率波动的时空分布特性(Spatiotemporal distribution characteristic of wind power fluctuation)[J].电网技术(Power System Technology),2011,35(2):110-114.
[2]陈国初,公维祥,冯兆红(Chen Guochu,Gong Weixiang,Feng Zhaohong).熵判别人工蜂群算法优化的风电功率组合预测模型(Combined wind power prediction model optimized by entropy criterion artificial bee colony algorithm)[J].电力系统及其自动化学报(Proceedings of the CSU-EPSA),2017,29(7):41-45,67.
[3]师洪涛,杨静玲,丁茂生,等(Shi Hongtao,Yang Jingling,Ding Maosheng,et al).基于小波-BP神经网络的短期风电功率预测方法(A short-term wind power prediction method based on wavelet decomposition and BP neural network)[J].电力系统自动化(Automation of Electric Power Systems),2011,35(16):44-48.
[4]叶林,刘鹏(Ye Lin,Liu Peng).基于经验模态分解和支持向量机的短期风电功率组合预测模型(Combined model based on EMD-SVM for short-term wind power prediction)[J].中国电机工程学报(Proceedings of the CSEE),2011,31(31):102-108.
[5]Bouffard F,Galiana F D.Stochastic security for operations planning with significant wind power generation[J].IEEETrans on Power Systems,2008,23(2):306-316.
[6]杨秀媛,肖洋,陈树勇(Yang Xiuyuan,Xiao Yang,Chen Shuyong).风电场风速和发电功率预测研究(Wind speed and generated power forecasting in wind farm)[J].中国电机工程学报(Proceedings of the CSEE),2005,25(11):1-5.
[7]茆美琴,周松林,苏建徽(Mao Meiqin,Zhou Songlin,Su Jianhui).基于脊波神经网络的短期风电功率预测(Short-term wind power forecast based on ridgelet neural network)[J].电力系统自动化(Automation of Electric Power Systems),2011,35(7):70-74.
[8]周松林,茆美琴,苏建徽(Zhou Songlin,Mao Meiqin,Su Jianhui).基于主成分分析与人工神经网络的风电功率预测(Prediction of wind power based on principal component analysis and artificial neural network)[J].电网技术(Power System Technology),2011,35(9):128-132.
[9]丁志勇,杨苹,杨曦,等(Ding Zhiyong,Yang Ping,Yang Xi,et al).基于连续时间段聚类的支持向量机风电功率预测方法(Wind power prediction method based on sequential time clustering support vector machine)[J].电力系统自动化(Automation of Electric Power Systems),2012,36(14):131-135,149.
[10]周松林,茆美琴,苏建徽(Zhou Songlin,Mao Meiqin,Su Jianhui).风电功率短期预测及非参数区间估计(Shortterm forecasting of wind power and non-parametric confidence interval estimation)[J].中国电机工程学报(Proceedings of the CSEE),2011,31(25):10-16.
[11]李俊芳,张步涵,谢光龙,等(Li Junfang,Zhang Buhan,Xie Guanglong,et al).基于灰色模型的风速-风电功率预测研究(Grey predictor models for wind speed-wind power prediction)[J].电力系统保护与控制(Power System Protection and Control),2010,38(19):151-159.
[12]Papaefthymiou G,Kl?ckl B.MCMC for wind power simulation[J].IEEE Trans on Energy Conversion,2008,23(1):234-240.
[13]Morales J R,Minguez R,Conejo A J.A methodology to generate statistically dependent wind speed scenarios[J].Applied Energy,2010,87(3):843-855.
[14]Chen P,Pedersen T,Bak-Jensen B,et al.ARIMA-based time series model of stochastic wind power generation[J].IEEE Trans on Power Systems,2010,25(2):667-676.
[15]吴怀宇.时间序列分析与综合[M].武汉:武汉大学出版社,2004.
[16]丁义明,范文涛,谭秋衡,等(Ding Yiming,Fan Wentao,Tan Qiuheng,et al).数据流的非平稳性度量(Nonstationarity measure of data stream)[J].数学物理学报(Acta Mathematica Scientia),2010,30(5):1364-1376.
[17]张世英,樊智,郭名媛.协整理论与波动模型[M].北京:清华大学出版社,2013.
[18]司守奎,孙玺菁.数学建模算法与应用[M].北京:国防工业出版社,2014.
[19]冬雷,王丽婕,高爽,等(Dong Lei,Wang Lijie,Gao Shuang,et al).基于混沌时间序列的大型风电场发电功率预测建模与研究(Modeling and analysis of prediction of wind power generation in the large wind farm based on chaotic time series)[J].电工技术学报(Transactions of China Electrotechnical Society),2008,23(12):125-129.
[2]陈国初,公维祥,冯兆红(Chen Guochu,Gong Weixiang,Feng Zhaohong).熵判别人工蜂群算法优化的风电功率组合预测模型(Combined wind power prediction model optimized by entropy criterion artificial bee colony algorithm)[J].电力系统及其自动化学报(Proceedings of the CSU-EPSA),2017,29(7):41-45,67.
[3]师洪涛,杨静玲,丁茂生,等(Shi Hongtao,Yang Jingling,Ding Maosheng,et al).基于小波-BP神经网络的短期风电功率预测方法(A short-term wind power prediction method based on wavelet decomposition and BP neural network)[J].电力系统自动化(Automation of Electric Power Systems),2011,35(16):44-48.
[4]叶林,刘鹏(Ye Lin,Liu Peng).基于经验模态分解和支持向量机的短期风电功率组合预测模型(Combined model based on EMD-SVM for short-term wind power prediction)[J].中国电机工程学报(Proceedings of the CSEE),2011,31(31):102-108.
[5]Bouffard F,Galiana F D.Stochastic security for operations planning with significant wind power generation[J].IEEETrans on Power Systems,2008,23(2):306-316.
[6]杨秀媛,肖洋,陈树勇(Yang Xiuyuan,Xiao Yang,Chen Shuyong).风电场风速和发电功率预测研究(Wind speed and generated power forecasting in wind farm)[J].中国电机工程学报(Proceedings of the CSEE),2005,25(11):1-5.
[7]茆美琴,周松林,苏建徽(Mao Meiqin,Zhou Songlin,Su Jianhui).基于脊波神经网络的短期风电功率预测(Short-term wind power forecast based on ridgelet neural network)[J].电力系统自动化(Automation of Electric Power Systems),2011,35(7):70-74.
[8]周松林,茆美琴,苏建徽(Zhou Songlin,Mao Meiqin,Su Jianhui).基于主成分分析与人工神经网络的风电功率预测(Prediction of wind power based on principal component analysis and artificial neural network)[J].电网技术(Power System Technology),2011,35(9):128-132.
[9]丁志勇,杨苹,杨曦,等(Ding Zhiyong,Yang Ping,Yang Xi,et al).基于连续时间段聚类的支持向量机风电功率预测方法(Wind power prediction method based on sequential time clustering support vector machine)[J].电力系统自动化(Automation of Electric Power Systems),2012,36(14):131-135,149.
[10]周松林,茆美琴,苏建徽(Zhou Songlin,Mao Meiqin,Su Jianhui).风电功率短期预测及非参数区间估计(Shortterm forecasting of wind power and non-parametric confidence interval estimation)[J].中国电机工程学报(Proceedings of the CSEE),2011,31(25):10-16.
[11]李俊芳,张步涵,谢光龙,等(Li Junfang,Zhang Buhan,Xie Guanglong,et al).基于灰色模型的风速-风电功率预测研究(Grey predictor models for wind speed-wind power prediction)[J].电力系统保护与控制(Power System Protection and Control),2010,38(19):151-159.
[12]Papaefthymiou G,Kl?ckl B.MCMC for wind power simulation[J].IEEE Trans on Energy Conversion,2008,23(1):234-240.
[13]Morales J R,Minguez R,Conejo A J.A methodology to generate statistically dependent wind speed scenarios[J].Applied Energy,2010,87(3):843-855.
[14]Chen P,Pedersen T,Bak-Jensen B,et al.ARIMA-based time series model of stochastic wind power generation[J].IEEE Trans on Power Systems,2010,25(2):667-676.
[15]吴怀宇.时间序列分析与综合[M].武汉:武汉大学出版社,2004.
[16]丁义明,范文涛,谭秋衡,等(Ding Yiming,Fan Wentao,Tan Qiuheng,et al).数据流的非平稳性度量(Nonstationarity measure of data stream)[J].数学物理学报(Acta Mathematica Scientia),2010,30(5):1364-1376.
[17]张世英,樊智,郭名媛.协整理论与波动模型[M].北京:清华大学出版社,2013.
[18]司守奎,孙玺菁.数学建模算法与应用[M].北京:国防工业出版社,2014.
[19]冬雷,王丽婕,高爽,等(Dong Lei,Wang Lijie,Gao Shuang,et al).基于混沌时间序列的大型风电场发电功率预测建模与研究(Modeling and analysis of prediction of wind power generation in the large wind farm based on chaotic time series)[J].电工技术学报(Transactions of China Electrotechnical Society),2008,23(12):125-129.