摘要
具有分布广、蕴藏量大的风能,作为一种清洁环保的可再生能源越来越受到重视。在大力发展风力发电的同时,风电入网对电能质量、电力系统安全运行以及电力供需平衡也带来了挑战。如何降低风电入网给电力系统带来的影响、以及提高风力发电的竞争力,是发展风电技术一个急需解决的难题。对风电功率进行短期预测,其精确的预测结果能够指导电力调度部门及时调整调度计划、减少备用容量、降低运行成本,是促进风电技术发展有效途径之一。
本文主要是对风速及风电功率短期预测技术进行了研究,利用风速混沌特性进行风速短期预测,然后在风电功率曲线的基础上完成对风电功率的预测。对风速时间序列的属性进行正确的分析,将有利于科学合理的建立风速短期预测模型。先对风速时间序列进行相空间重构,通过改进C-C法与原C-C法相比,利用改进C-C方法的高可靠性、计算速度快并且能同时估算出嵌入维数和延迟时间的特点估算出风速时间序列的重构参数。分别计算出风速时间序列的关联维数和最大Lyapunov指数,验证风速时间序列的混沌特性,将混沌相空间重构引入到风速预测中,为风速短期预测奠定了理论基础。利用加权零阶局域法与加权一阶局域法两种常用传统的混沌时间序列预测方法对风速进行短期预测,预测结果表明了利用混沌时间序列预测方法对风速进行预测的可行性。着重介绍了一种具有高精度混沌时间序列预测的新型递归神经网络——回声状态网络,对其网络的建立过程、基于混沌时间序列的迭代预测和直接预测进行了详细的分析,将风速混沌时间序列和回声状态网络相结合,建立了风速单步预测和提前8小时的风速短期预测模型,MATLAB的仿真结果表明了该预测方法的有效性,与传统混沌时间序列预测方法相比,提高了预测精度。在标准风电功率曲线的基础上,通过风速短期预测值得到了风电功率的预测值,仿真结果表明由于标准的风电功率曲线不完全符合实际风速与风电功率的关系,使得风电功率预测误差大于风速的预测误差,但预测精度还算令人满意。本文所提出的预测方法为风电功率预测技术开辟了一个新的空间。
With wide distribution and large reservation, wind as a environmentally friendly renewable energy is received more and more attention. In the development of wind power, the wind power grid is also brought challenge to power quality, the safe operation of the power system and power supply and demand balance. How to reduce the impact of wind power network to the power system, as well as enhance the competitiveness of wind power generation is a difficult problem of developing wind power technology, which is in need of solution. The power dispatching department can be guided by the accurate predictions of short-term wind power prediction to timely adjust scheduling plan, so as to reduce the reserve capacity and operation costs, which is an effective way to promote the development of wind power technology.
In this paper, wind speed and wind power short-term prediction technology is studied, through the short-term forecast for wind speed, the wind power forecast is completed based on the wind power curve. Correct analysis of wind speed time series properties will be beneficial to build the short-term prediction model scientifically and rationally. First, phase space reconstruction of wind speed time series is taken. By comparing of the original C-C method, the improved C-C method has the characteristics of high reliability, fast calculation and can also estimate the embedding dimension and delay time at the same time, so it is chose to estimate the parameters of the reconstruction of wind speed time series. Then, the chaotic characteristics of wind speed time series is verified by the correlation dimension and maximum Lyapunov index, so chaotic phase space reconstruction can be introduced to the wind speed prediction and laid the foundation for the prediction. And two commonly used traditional chaotic time series prediction method, the weighted zero-order local low and the weighted one-order local law is used to predict the short-term wind speed. The prediction results show that using chaotic time series prediction method to predict the short-term wind speed is feasibility. Finally, a new type of recurrent neural networks—echo state network which has high precision of chaotic time series prediction is introduced, and the process of establish the network, chaotic time series based iterative prediction and direct prediction are analyzed in detail. Combining the wind speed chaotic time series and the echo state network, the forecasting model of single-step wind speed prediction and eight hours ahead wind speed short-term prediction are established, and the MATLAB simulation results show that the prediction method is valid. Compared to the traditional chaotic time series prediction method, the prediction accuracy is improved. On the basis of standard wind power curve, the prediction values of wind power are obtained by the prediction values of wind speed. The simulation results indicate that the standard power curve is not fully comply with the actual relationship between wind speed and wind power, so as to make wind power prediction error is greater than the wind speed prediction error, but the prediction accuracy is still satisfactory. The proposed prediction method has opened a new space for wind power prediction technology.
引文
[1]韩爽.风电场功率短期预测方法研究[D]：[博士学位论文].北京：华北电力大学.2008.
[2]杨志凌.风电场功率短期预测方法优化的研究[D]：[博士学位论文].北京：华北电力大学.2011.
[3]李叶,叶林.基于改进持续法的短期风电功率预测[J].农业工程学报,2010,26(12)：182-187.
[4]丁明,张立军,吴义纯.基于时间序列分析的风电场风速预测模型[J].电力自动化设备,2005,25(8)：32-34.
[5]潘迪夫.刘辉.李燕吃.基于时间序列分析和卡尔曼滤波算法的风电场风速预测优化模型[J].电网技术.2008,32(7)：82-86.
[6]刘辉,田红旗,CHEN Chao,等.基于小波分析法与神经网络法的非平稳风速信号短期预测优化算法[J].中南大学学报,2011,42(9)：2704-2711.
[7]师洪涛,杨静玲,丁茂生,等.基于小波-BP神经网络的短期风电功率预测方法[J].电力系统自动化,2011,35(16)：44-48.
[8]王世谦,苏娟,杜松怀,等.基于小波变换和神经网络的短期风电功率预测方法[J].农业工程学报,2010,26(2)：125-129.
[9]黄磊.基于动态神经网络的风电场输出功率预测及其应用的研究[D]:[硕士学位论文].哈尔滨：哈尔滨工业大学.2011.
[10]张国强,张伯明.基于组合预测的风电场风速及风电机功率预测[J].电力自动化,2009,23(18).
[11]Bates J M,Granger C W J. The combination of forecast [J].Operational Research Quarterly. 1969,20(4):451-468
[12]Bossanyi E A. Short-term wind prediction using Kalman filters [J].Wind Engineering. 1985,9(1):1-8.
[13]Landberg L. Predikior: an on-line Prediction system [C].Wind Power for the 21st Century, EUWEC Special Topic Conference,Kassel,Germany,2000.
[14]Nielsen T S. Madsen H. WPPT: a tool for wind Power Prediction [C]. Proceedings of Wind Power for the 21st Century,Kassel,Germany,2000.
[15]Giebel G, Landberg L, Joensen Alfred K, et al. The zephyr-Project:the next generation Prediction system [C]. Proceedings of Wind Power for the 21st Century, Kassel, Germany, 2000.
[16]Zack J W, Brower M C, Bailey B H. Validating of the forewind model in wind forecasting Application [C]. Wind Power for the 21st Century, EUWEC Special Topic Conference, Kassel,Germany,2000.
[17]Lange M, Focken U, Heinemann D. Previento-regional wind Power Prediction with risk Control [C].Proceedings of the World Wind Energy Conference,Berlin,2002.
[18]Lange B, Rohrig K, Ernst B, et al. Wind Power Prediction in Germany:recent advances and future challenges [C],European Wind Energy Conference,Athens,2006.
[19]Sanchez I, Usaola O R J, Velaseo C, et al.Sipreolico:a wind Power prediction system based on flexible combination of dynamic models. Application to the Spanish power system [C].First IEAJoint Action Symposium on Wind Forecasting Techniques, NorrkoPing, Sweden,2002.
[20]Hatziargyriou N, Contaxi G, Matos M, et al. More care:Advice for secure operation of isolated Power systems with increased renewable energy penetration and storage [C]. European Wind Energy Conference,CoPenhagen,2001.
[21]张华,曾杰.基于支持向量机的风速预测模型研究[J].太阳能学报,2010,31(7)：928-932.
[22]曾杰,张华.基于最小二乘支持向量机的风速预测模型[J].电网技术,2009,33(18)：144-147.
[23]杨绍清,贾传荧.两种实用的相空间重构方法[J].物理学报,2002,51(11)：2452-2458.
[24]冬雷,王丽婕,高爽,等.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[J].电工技术学报,2008,23(12)：125-129.
[25]陶小创,樊焕贞,吕琛,等.基于小波神经网络的混沌时间序列预测及应用[J].南京航空航天大学学报,2011,43(s)：174-178.
[26]吕金虎,陆君安,陈士华.混沌时间序列分析及其应用[M].武汉：武汉大学出版社.2002.
[27]秦奕青,蔡卫东,杨炳儒.非线性时间序列的相空间重构技术研究[J].系统仿真学报,2008,20(11)：2969-2973.
[28]陆振波,蔡志明,姜可宇.基于改进的C-C方法的相空间重构参数选择[J].系统仿真学报,2007,19(11)：2527-2529,2538.
[29]陶佳,张宏,朱国荣,等.基于优化相空间重构技术的风电场发电功率预测研究[J].中国电机工程学报,2011,31(28)：9-14.
[30]王东风,张有玥,韩璞,等.风电场风速时间序列的复杂动力学特性分析[J].同济大学学报(自然科学版),2010,38(12)：1828-1831.
[31]王媛媛.基于BP神经网络的混沌时间序列预测方法研究[D]：[硕士学位论文].石家庄：河北经贸大学.2011.
[32]Famer J D, Sidorowich J J. Predicting chaotic time series [J].Physical Review Letters,1987, 59 (8):845-848.
[33]谢忠玉.电力短期负荷时间序列混沌特性分析[D]:[博士学位论文].哈尔滨：哈尔滨工程大学.2010.
[34]Giona M, Cimagalli V, Morgavi G, et al. Local prediction of chaotic time series [J]. Proceedings of IEEE workshop on nonlinear signal and image proceeding,1991,894-897.
[35]Han M, Xi J, Xu S, et al. Prediction of chaotic time series based on recurrent predictior neural network [J].IEEE Transactiona on Signal Processing,2004,52(12):3409-3416.
[36]Cowper M R, Mulgrew B, Unsworth C P. Nonlinear prediction of chaotic signals using a normalized radial basis function network [J].Signal Procedding,2002,82 (5):775-789.
[37]H Jeager.The"echo state"Approach to Analyzing and Training Recurrent Neural Networks [R]. GMD Report 148,GMD-German National Research Institute for Computer Science, 2001.
[38]史志伟,韩敏.ESN岭回归学习算法及混沌时间序列预测[J].控制与决策,2007,22(3)：258-261,267.
[39]李德才,韩敏.基于鲁棒回声状态网络的混沌时间序列预测研究[J].物理学报,2011,60(10)：1-8.
[40]张亚琦.基于改进回声状态网络的混沌时间序列多步预测[D]：[硕士学位论文].兰州：兰州大学.2010.
[41]韩敏,史志伟,郭伟.储备池状态空间重构与混沌时间序列预测[J].物理学报,2007,56(1)：43-50.
[42]韩敏,穆大芸.回声状态网络LM算法及混沌时间序列预测[J].控制与决策,2011,26(10)：1469-1472,1478.
[43]郭阳明,孙姜燕,付琳娟,等.基于回声状态网络和主元分析的混沌时间序列预测模型[J].西北工业大学学报,2010,28(6)：946-951.
[44]米增强,刘兴杰,张艳青,等.基于混沌分析和神经网络的风速直接多步预测[J].太阳能学报,2011,32(6)：901-906.
[45]H Jaeger,H Haas.Hamessing nonlinearity:Predicting chaotic systems and saving energy in wireless telecommunication [J].Science,2004,308:78-80.
[46]刘兴杰.风电输出功率预测方法与系统[D]：[博士学位论文].北京：华北电力大学.2010.