基于时间序列分析的大型风电场功率预测方法研究
|
摘要
风电功率预测对电力系统有着重要的意义,精确的预测可以降低风电并网带来的冲击,提高风电渗透率,增强电力系统运行的稳定性。由于风电的随机性和波动性,加之我国风电场所处地形复杂,风况相对不稳定,我国的风电预测在精度上仍然有待提高。
在此背景下,本文以时间序列分析为基础,选取多种风电预测模型对风电场风速和功率进行短期预测,主要工作包括以下几个方面：
对风电场运行数据进行统计分析,并对其进行预处理。分别对多种方法进行研究,建立了风电预测的单一模型和综合模型。
提出了基于遗忘因子优化的AR参数估计模型,将系统辨识中比较常用且精度较高的最小二乘估计应用于风电功率预测,利用递推方程进行建模。为避免数据饱和现象,在最小二乘算法中加入了遗忘因子来控制历史数据对未来预测情况的影响,其中遗忘因子根据误差情况进行自动调整来控制数据窗的大小
提出了基于D-S证据理论的改进ARMA模型,将各单一模型的预测值作为多方面信息进行融合,以达到修正ARMA参数的目的。之后,对基于D-S证据理论的改进ARMA模型在选型方面进行了优化。
在已有的优化模型的基础上,提出了基于组合模型的递推预测算法,将各种算法进行误差验证,选出误差最小的模型进行下一步预测,再利用变遗忘因子最小二乘算法对结果进行修正,使预测误差进一步降低。
对内蒙古某风电场功率进行了预测,首先对单机风速进行预测,然后对各台风机进行功率拟合,最后将单机功率叠加得到风电场总功率。将各种算法的预测结果进行比较,实验证明组合模型在预测中具有较好的精度。
Wind power prediction is significant to electrical power system. An accurate forecasting can reduce the impact of wind power on power system, and increase the penetration of wind power and enhance the stability of power system operation. Because of the intermittence and fluctuation of the wind and the complexity of China's topography, the wind condition is not stable relatively. So China's wind power prediction is remained to be improved in accuracy.
Under the circumstances, several models are studied and used to make the prediction based on Time Series Analysis in this paper. The main works are as follow:
Statistical analysis of wind farm operation data is studied. Several single models and integrated models are established and the preprocessing is made before the establishment of models.
An integrated model based on D-S Evidence Theory is put forward. The weights of single models are extracted as the evidence and the integrated model is established with multi-fusion of the belief function before the final result is presented. Then, the integrated model is optimized in the selection stage of single models.
A recursive least square model based on auto-regulation forgetting factor is put forward. The result is predicted by recursive equation, and the auto-regulation forgetting factor is designed to adjust automatically according to the error.
The wind power prediction model library is built based on optimized models and the optimization algorism of recursive prediction based on variable models is presented. The least relative error model is selected to predict wind speed or power, and the result is revised by recursive least square model based on auto-regulation forgetting factor, at least, the final result is more accurate.
在此背景下,本文以时间序列分析为基础,选取多种风电预测模型对风电场风速和功率进行短期预测,主要工作包括以下几个方面：
对风电场运行数据进行统计分析,并对其进行预处理。分别对多种方法进行研究,建立了风电预测的单一模型和综合模型。
提出了基于遗忘因子优化的AR参数估计模型,将系统辨识中比较常用且精度较高的最小二乘估计应用于风电功率预测,利用递推方程进行建模。为避免数据饱和现象,在最小二乘算法中加入了遗忘因子来控制历史数据对未来预测情况的影响,其中遗忘因子根据误差情况进行自动调整来控制数据窗的大小
提出了基于D-S证据理论的改进ARMA模型,将各单一模型的预测值作为多方面信息进行融合,以达到修正ARMA参数的目的。之后,对基于D-S证据理论的改进ARMA模型在选型方面进行了优化。
在已有的优化模型的基础上,提出了基于组合模型的递推预测算法,将各种算法进行误差验证,选出误差最小的模型进行下一步预测,再利用变遗忘因子最小二乘算法对结果进行修正,使预测误差进一步降低。
对内蒙古某风电场功率进行了预测,首先对单机风速进行预测,然后对各台风机进行功率拟合,最后将单机功率叠加得到风电场总功率。将各种算法的预测结果进行比较,实验证明组合模型在预测中具有较好的精度。
Wind power prediction is significant to electrical power system. An accurate forecasting can reduce the impact of wind power on power system, and increase the penetration of wind power and enhance the stability of power system operation. Because of the intermittence and fluctuation of the wind and the complexity of China's topography, the wind condition is not stable relatively. So China's wind power prediction is remained to be improved in accuracy.
Under the circumstances, several models are studied and used to make the prediction based on Time Series Analysis in this paper. The main works are as follow:
Statistical analysis of wind farm operation data is studied. Several single models and integrated models are established and the preprocessing is made before the establishment of models.
An integrated model based on D-S Evidence Theory is put forward. The weights of single models are extracted as the evidence and the integrated model is established with multi-fusion of the belief function before the final result is presented. Then, the integrated model is optimized in the selection stage of single models.
A recursive least square model based on auto-regulation forgetting factor is put forward. The result is predicted by recursive equation, and the auto-regulation forgetting factor is designed to adjust automatically according to the error.
The wind power prediction model library is built based on optimized models and the optimization algorism of recursive prediction based on variable models is presented. The least relative error model is selected to predict wind speed or power, and the result is revised by recursive least square model based on auto-regulation forgetting factor, at least, the final result is more accurate.
引文
[2]Tande J. Grid integration of wind farms[J]. Wind Energy, 2003, 6:281-295
[3]李建林.风电规模化发展,离不开风电功率预测[J].变频器世界,2010,4,49-51
[4]彭晖.风电场风电量短期预测技术研究.东南大学,2005
[5]何育.基于非经典计量经济学的短期风电功率预测.江苏,东南大学,2010
[6]杨志凌.风电场功率短期预测方法优化的研究.北京,华北电力大学,2011
[7]贺电.大型风电场短期功率预测研究.北京,北京交通大学,2011
[8]Billinton R, HUA Chen, Ghajar R, A Sequential Ernst B, Oakleaf B, Ahlstrom M L, et al. Predicting the Wind[J]. IEEE Power&Energy Magazine. 2007,11:79-89
[9]周双喜,鲁宗相.风力发电与电力系统[M].北京：中国电力出版社,2011
[10]吴石林,张玘.误差分析与数据处理[M].北京：清华大学出版社,2010
[11]阎晓明,李言俊,陈新海.一种自动调整遗忘因子的快速时变参数辨识方法[J].自动化学报,1991.5,17(3)：336-339
[12]储鹏鹏.基于小波变换的图像去噪方法研究.陕西：西安电子科技大学,2009
[13]陈雪松.基于递归最小二乘法的电压闪变研究[J].四川：四川大学,2006
[14]白鹏等.支持向量机理论及工程应用实例[M].陕西：西安电子科技大学出版社,2008
[15]夏定纯,徐涛.人工智能技术与方法[M].湖北：华中科技大学出版社,2004
[16]杨秀媛,肖洋,陈树勇.风电场风速和发电功率预测研究[J].中国电机工工程学报,2005,25(11)：1-5.
[17]吴京秋,孙奇,杨伟,杨杰.基于D-S证据理论的短期负荷预测模型融合[J].电力自动化设备,2009,29(4)：66-70
[18]X. Wang, G. Sideratos, N. Hatziargyriou, L. H. Tsoukalas. Wind speed forecasting for power system operational planning.
[19]朱成骥,孙宏斌,张伯明.基于最大信息熵原理的短期负荷预测综合模[J].中国电机工程学报,2005,25(19)：1-6
[20]韩爽.风电功率短期预测方法研究[D].北京：华北电力大学,2008
[21]Weiqing Cao, Pan Fu and Weilin Li. The Diagnosis of Tool Wear Based on RBF Neural Networks and D-S Evidence Theory.
[22]倪明,高晓萍,单渊达.证据理论在中期负荷预测中的应用[J].中国电机工程学报,1997,17(3)：199-203
[23]杨风暴,王肖霞.D-S证据理论的冲突证据合成方法[M].北京：国防工业出版社,2010
[24]Li YiBo, Wang Ning. Based on D-S Evidence Theory of Information Fusion Improved Method.
[25]The Quality of a 48-Hours Wind Power Forecast Using the German and Danish Weather prediction Model. Riso National Laboratory, P.O. box,49.
[26]Xia Dong, Wu Junyong. Study on the Integrated Model for Short-term Wind Power Forecasting Based on D-S Evidence Theory. The 4th Asia-Pacific Power and Energy Engineering Conference (APPEEC 2012). Shanghai,2012-3
[27]姚锡凡,李曼.人工智能技术及应用[M].北京：中国电力出版社,2008
[28]李鹏波,胡德文,张纪阳,谢红卫.系统辨识[M].北京：中国水利水电出版社,2010
[29]杨叔子,吴雅,轩建平.时间序列分析的工程应用(上、下)[M].湖北：华中科技大学出版社,2007
[30]王军选,田小平,曹红梅.信息论基础与编码[M].北京：人民邮电出版社,2011
[31]潘立登,潘仰东.先进控制系统应用于维护[M].北京：中国电力出版社,2010
[32]曹承志等.人工智能技术[M].北京：清华大学出版社,2010
[33]李倩茹,王于丁,张晓芳.一种变遗忘因子RLS算法的分析与仿真[J].现代电子技术,2008-17：45-47.
[34]A. Osorio Cordero, D.Q. Mayne. Deterministic convergence of a self-tuning regulator with variable forgetting factor. IEE PROC, 1981,1,Vol.128(1):19-23
[35]Costa A, Crespo A, Navarro J, et al. A review on the young history of the wind power short-term prediction, RENEWABLE & SUSTAINABLE ENERGY REVIEWS, Aug 2008.12(6). pp1725-1744.
[36]王燕.应用时间序列分析[M].北京：中国人民大学出版社,2009
[37]The Quality of a 48-Hours Wind Power Forecast Using the German and Danish Weather prediction Model. Riso National Laboratory, P.O. box,49.
[38]谷兴凯,范高峰,王晓蓉.风电功率预测技术综述,现代电力,2007/12,31(2),pp335-338.
[39]潘迪夫,刘辉,李燕飞.风电场风速短期多步预测改进算法[J].中国电机工程学报2008,28(26)：87-91.
[40]石峰,莫忠息.信息论基础.武汉：武汉大学出版社,2006.
[41]He Dian, Wu Jun-yong, Zhang Xi-lu, Ji Lu-yu, Li Xue. An Integrated Model for Wind Power Forecasting Based on Maximum Entropy Principle. ICEEAC2010.
[42]李伯颐.基于模糊聚类预处理的风电场短期风速和发电功率预测.上海：上海交通大学,2009
[43]张国忠,汪世汉.自动调整遗忘因子的LKL辨识方法在销售预测中的应用[J].预测,1996,3,63-64
[44]袁晶,张小峰.基于遗忘因子的BP神经网络水文实时预报方法[J].水科学进展2004,15(6)：787-792
[45]刘国海,吕汉闻,陈兆岭等.基于修正遗忘因子RLS算法的谐波电流检测新方法[J].电工技术学报,2010,25(1)：172-177.
[46]杨金明,王伟强.一种可变步长LMS算法及其性能分析[J].华南理工大学学报(自然科学版),2006,34(4)：61-64.
[47]向炳新.针对电信行业自适应智能话务预测模型的研究与实现[D].2007
[48]陈雪松,杨洪耕.递归最小二乘法追踪电压闪变包络线[J].电力自动化设备,2006,26(6)：30-32.
[49]吕振肃,黄石.一种改进的变步长ELMS算法[J].电子与信息学报,2005,27(10)：1524-1526
[50]李竹,杨培林,行小帅等.一种改进变步长LMS算法及其在系统辨识中的应用[J].仪器仪表,2007,28(7)：1340-1344.
[51]刘聪,孙秀霞,李海军等.带遗忘因子的限定记忆辨识算法[J].电光与控制,2006,13(1)：48-49
[52]李素芳,李颖,闫智义等.单通道RLS算法自适应噪声消除及其遗忘因子的确定[J].长春工业大学学报(自然科学版),2007,28(2)：224-228.
[53]牛东晓,曹树华,赵磊,张文文.电力负荷预测技术及其应用.北京：中国电力出版社,1998
[54]The Quality of a 48-Hours Wind Power Forecast Using the German and Danish Weather prediction Model. Riso National Laboratory, P.O. box,49.
[3]李建林.风电规模化发展,离不开风电功率预测[J].变频器世界,2010,4,49-51
[4]彭晖.风电场风电量短期预测技术研究.东南大学,2005
[5]何育.基于非经典计量经济学的短期风电功率预测.江苏,东南大学,2010
[6]杨志凌.风电场功率短期预测方法优化的研究.北京,华北电力大学,2011
[7]贺电.大型风电场短期功率预测研究.北京,北京交通大学,2011
[8]Billinton R, HUA Chen, Ghajar R, A Sequential Ernst B, Oakleaf B, Ahlstrom M L, et al. Predicting the Wind[J]. IEEE Power&Energy Magazine. 2007,11:79-89
[9]周双喜,鲁宗相.风力发电与电力系统[M].北京：中国电力出版社,2011
[10]吴石林,张玘.误差分析与数据处理[M].北京：清华大学出版社,2010
[11]阎晓明,李言俊,陈新海.一种自动调整遗忘因子的快速时变参数辨识方法[J].自动化学报,1991.5,17(3)：336-339
[12]储鹏鹏.基于小波变换的图像去噪方法研究.陕西：西安电子科技大学,2009
[13]陈雪松.基于递归最小二乘法的电压闪变研究[J].四川：四川大学,2006
[14]白鹏等.支持向量机理论及工程应用实例[M].陕西：西安电子科技大学出版社,2008
[15]夏定纯,徐涛.人工智能技术与方法[M].湖北：华中科技大学出版社,2004
[16]杨秀媛,肖洋,陈树勇.风电场风速和发电功率预测研究[J].中国电机工工程学报,2005,25(11)：1-5.
[17]吴京秋,孙奇,杨伟,杨杰.基于D-S证据理论的短期负荷预测模型融合[J].电力自动化设备,2009,29(4)：66-70
[18]X. Wang, G. Sideratos, N. Hatziargyriou, L. H. Tsoukalas. Wind speed forecasting for power system operational planning.
[19]朱成骥,孙宏斌,张伯明.基于最大信息熵原理的短期负荷预测综合模[J].中国电机工程学报,2005,25(19)：1-6
[20]韩爽.风电功率短期预测方法研究[D].北京：华北电力大学,2008
[21]Weiqing Cao, Pan Fu and Weilin Li. The Diagnosis of Tool Wear Based on RBF Neural Networks and D-S Evidence Theory.
[22]倪明,高晓萍,单渊达.证据理论在中期负荷预测中的应用[J].中国电机工程学报,1997,17(3)：199-203
[23]杨风暴,王肖霞.D-S证据理论的冲突证据合成方法[M].北京：国防工业出版社,2010
[24]Li YiBo, Wang Ning. Based on D-S Evidence Theory of Information Fusion Improved Method.
[25]The Quality of a 48-Hours Wind Power Forecast Using the German and Danish Weather prediction Model. Riso National Laboratory, P.O. box,49.
[26]Xia Dong, Wu Junyong. Study on the Integrated Model for Short-term Wind Power Forecasting Based on D-S Evidence Theory. The 4th Asia-Pacific Power and Energy Engineering Conference (APPEEC 2012). Shanghai,2012-3
[27]姚锡凡,李曼.人工智能技术及应用[M].北京：中国电力出版社,2008
[28]李鹏波,胡德文,张纪阳,谢红卫.系统辨识[M].北京：中国水利水电出版社,2010
[29]杨叔子,吴雅,轩建平.时间序列分析的工程应用(上、下)[M].湖北：华中科技大学出版社,2007
[30]王军选,田小平,曹红梅.信息论基础与编码[M].北京：人民邮电出版社,2011
[31]潘立登,潘仰东.先进控制系统应用于维护[M].北京：中国电力出版社,2010
[32]曹承志等.人工智能技术[M].北京：清华大学出版社,2010
[33]李倩茹,王于丁,张晓芳.一种变遗忘因子RLS算法的分析与仿真[J].现代电子技术,2008-17：45-47.
[34]A. Osorio Cordero, D.Q. Mayne. Deterministic convergence of a self-tuning regulator with variable forgetting factor. IEE PROC, 1981,1,Vol.128(1):19-23
[35]Costa A, Crespo A, Navarro J, et al. A review on the young history of the wind power short-term prediction, RENEWABLE & SUSTAINABLE ENERGY REVIEWS, Aug 2008.12(6). pp1725-1744.
[36]王燕.应用时间序列分析[M].北京：中国人民大学出版社,2009
[37]The Quality of a 48-Hours Wind Power Forecast Using the German and Danish Weather prediction Model. Riso National Laboratory, P.O. box,49.
[38]谷兴凯,范高峰,王晓蓉.风电功率预测技术综述,现代电力,2007/12,31(2),pp335-338.
[39]潘迪夫,刘辉,李燕飞.风电场风速短期多步预测改进算法[J].中国电机工程学报2008,28(26)：87-91.
[40]石峰,莫忠息.信息论基础.武汉：武汉大学出版社,2006.
[41]He Dian, Wu Jun-yong, Zhang Xi-lu, Ji Lu-yu, Li Xue. An Integrated Model for Wind Power Forecasting Based on Maximum Entropy Principle. ICEEAC2010.
[42]李伯颐.基于模糊聚类预处理的风电场短期风速和发电功率预测.上海：上海交通大学,2009
[43]张国忠,汪世汉.自动调整遗忘因子的LKL辨识方法在销售预测中的应用[J].预测,1996,3,63-64
[44]袁晶,张小峰.基于遗忘因子的BP神经网络水文实时预报方法[J].水科学进展2004,15(6)：787-792
[45]刘国海,吕汉闻,陈兆岭等.基于修正遗忘因子RLS算法的谐波电流检测新方法[J].电工技术学报,2010,25(1)：172-177.
[46]杨金明,王伟强.一种可变步长LMS算法及其性能分析[J].华南理工大学学报(自然科学版),2006,34(4)：61-64.
[47]向炳新.针对电信行业自适应智能话务预测模型的研究与实现[D].2007
[48]陈雪松,杨洪耕.递归最小二乘法追踪电压闪变包络线[J].电力自动化设备,2006,26(6)：30-32.
[49]吕振肃,黄石.一种改进的变步长ELMS算法[J].电子与信息学报,2005,27(10)：1524-1526
[50]李竹,杨培林,行小帅等.一种改进变步长LMS算法及其在系统辨识中的应用[J].仪器仪表,2007,28(7)：1340-1344.
[51]刘聪,孙秀霞,李海军等.带遗忘因子的限定记忆辨识算法[J].电光与控制,2006,13(1)：48-49
[52]李素芳,李颖,闫智义等.单通道RLS算法自适应噪声消除及其遗忘因子的确定[J].长春工业大学学报(自然科学版),2007,28(2)：224-228.
[53]牛东晓,曹树华,赵磊,张文文.电力负荷预测技术及其应用.北京：中国电力出版社,1998
[54]The Quality of a 48-Hours Wind Power Forecast Using the German and Danish Weather prediction Model. Riso National Laboratory, P.O. box,49.